1{"work_id": 3134501457, "title": "KLASIFIKASI INDEKS PEMBANGUNAN MANUSIA KABUPATEN/KOTA DI INDONESIA MENGGUNAKANMETODE RANDOM FOREST", "abstract": "Keberhasilan suatu negara dapat dilihat dari kondisi masyarakatnya. Suatu negara\r\nperlu mengembangkan masyarakat, yaitu dengan cara pembangunan manusia.\r\nPembangunan manusia terdiri dari tiga komponen dasar yaitu umur panjang dan\r\nsehat; pengetahuan dan kehidupan yang layak. Beberapa indikator yang mewakili\r\nketiga komponen tersebut diringkas menjadi satu nilai tunggal yaitu Indeks\r\nPembangunan Manusia. Oleh karena itu, untuk Random Forest merupakan\r\npengembangan dari metode Classification And Regression Trees (CART) dengan\r\nmenambahkan proses bagging, yang digunakan adalah resampling bootstrap.\r\nMetode ini mampu meningkatkan kestabilan pohon klasifikasi yang terbentuk dan\r\nakurasi prediksi yang dihasilkan. Penelitian ini akan mengklasifikasikan IPM\r\nkabupaten/kota di Indonesia dengan jumlah 514 kabupaten/kota. Variabel respon\r\ndalam penelitian ini adalah indeks pembangunan manusia dan variabel\r\nprediktornya adalah angka harapan hidup, rata-rata lama sekolah, harapan lama\r\nsekolah, dan pengeluaran perkapita yang disesuaikan. Berdasarkan analisis\r\nmenggunakan Random Forest didapatkan hasil bahwa banyaknya pohon yang\r\nterpilih adalah 100 dengan m yang dicobakan adalah 2 dan variabel yang paling\r\nberpengaruh dalam kenaikan Indeks Pembangunan Manusia adalah Pendapatan\r\nPer Kapita dengan nilai kepentingan sebesar 29.48. Ketepatan hasil klasifikasi\r\nyang diukur menggunakan Total Accuracy Rate (1-APER) menghasilkan nilai\r\nakurasi sebesar 93.69%. \r\nKata Kunci: Indeks Pembangunan Manusia, Random Forest, Total Accuracy Rate\r\n(1-APER)."}2{"work_id": 5673758, "title": "A Novel Reconfiguration Scheme in Quantum-Dot Cellular Automata for Energy Efficient Nanocomputing", "abstract": "Quantum-Dot Cellular Automata (QCA) is currently being investigated as an alternative to CMOS technology. There has been extensive study on a wide range of circuits from simple logical circuits such as adders to complex circuits such as 4-bit processors. At the same time, little if any work has been done in considering the possibility of reconfiguration to reduce power in QCA devices. This work presents one of the first such efforts when considering reconfigurable QCA architectures which are expected to be both robust and power efficient. We present a new reconfiguration scheme which is highly robust and is expected to dissipate less power with respect to conventional designs. An adder design based on the reconfiguration scheme will be presented in this thesis, with a detailed power analysis and comparison with existing designs. In order to overcome the problems of routing which comes with reconfigurability, a new wire crossing mechanism is also presented as part of this thesis."}3{"work_id": 75685577, "title": "Electromagnetic Side-Channel Analysis for Hardware and Software Watermarking", "abstract": "With more and more ICs being used in sectors requiring confidentiality and integrity like payment systems, military, finance and health, there is a lot of concern in the security and privacy of ICs. The widespread adoption of Intellectual Property (IP) based designs for modern systems like system on chips has reduced the time to market and saved a lot of money for many companies. But this has also opened the gates for problems like product piracy, IP theft and fraud. It is estimated that billions of dollars are lost annually to illegal manufacturing of Integrated Circuits. A possible solution to this problem of IP theft is to insert small circuits which are like unique IDs that only the owner or the registered verifier will know and detect in case of any conflict. The circuits that are inserted are called watermarks and are in some cases kept very small so as to be hidden. In such cases, we would need detection schemes that work well even with very small watermarks. In this work, we use Electro-Magnetic (EM) based side-channels for the detection of watermarks. Since the 90s, Side-channel Analyses have attracted significant attention within the cryptographic community as they are able to obtain secret information from smart cards and ICs. The power side-channel analysis is a very powerful method but EM side-channels are very useful as they will not need a resistor in series to the power supply and just needs passive observation of the EM radiations emanated by the IC. This passive monitoring will be a big advantage in the case of automated watermark detection used by a verifier.\\nIn this work, we start with EM side-channel analysis on FPGA for smaller designs. We insert watermarks on a Micro-controller, Smartcard and an FPGA and detect these watermarks using EM side-channel information emanated from the Design under Test. We used environments with different levels of noise interference. We compare the watermarking application using EM side-channels and Power side-channels in these different setups. These watermarks are very small and are hard to attack or remove by an attacker through reverse engineering or side-channel information. Due to the robustness against such attacks and the easy access of EM side-channels when compared to power side-channels, the EM side-channel based watermarks will be a very good solution for the IP theft problem. EM side-channel based watermark detection supports automation which companies of IP cores can make use of. We also extended this work to EM Side-channel Trojans as the concepts are similar"}4{"work_id": 3156168805, "title": "Ultra-low vibration closed-cycle cryogenic surface-electrode ion trap apparatus", "abstract": "Trapped ions are one of the leading platforms for building scalable quantum information processors. Current ion traps run on less than 100 qubits. A useful quantum computer needs to operate thousands or even millions of qubits. If systems reach gate fidelities that allow for quantum error correction, gate errors below the physical fidelity can be reached, which is important for scaling. The most promising platform for scaling up ion-based quantum computers are surface electrode traps in connection with the so-called QCCD (Quantum Charge Coupled Device) architecture. In an evolution of the standard (laser-based) approach to quantum logic gates, gates can be performed by applying an oscillating magnetic near-field gradient to the ions. Current fidelities are about one order of magnitude away from reaching the fault tolerant threshold with no fundamental limitations in sight. Scaling QCCDs is impeded by heating rates of the ions as well as vacuum quality. Ions have to be stored for at least as long as it takes for an algorithm to run and subsequent readout. Collisions with background gas and heating can expel ions from the trap. The microwave near-field approach of driving gates benefits from ions that are trapped close to the trap surface. Unfortunately, heating becomes more prominent near the surface. A cryogenic environment reduces the background pressure by several orders of magnitude and the kinetic energy of gas molecules by two orders of magnitude. Heating rates in a cryogenic trap are about two orders of magnitude lower than in a room temperature trap with otherwise equal physical properties. To operate a QCCD processor, it is required to employ phase-stable Raman laser beams. The cryogenic design of the apparatus needs to take into account the resulting requirements for ultra-low vibration amplitudes. This is particularly relevant for closed-cycle cooling systems, which are desirable from an economic point of view compared to liquid cryostats that boil off helium, but which also feature moving mechanical parts that can introduce a serious amount of vibrations. Here, we present a vibration isolated closed cycle Gifford-McMahon cryocooler with record low vibration amplitudes of 29 nm (RMS=7.8 nm) in the vertical and 51 nm (RMS=13.5 nm) in the horizontal direction. It contains a self-sustained inner vacuum chamber that houses the trap. It features 100 DC connections, 10 of which are able to carry up to 1 A of current and 8 high frequency lines. At 5 K, we can apply continuous 3.4 W of microwave power continuously at 1 GHz to the trap electrodes without heating the system. This is about 5.8 times more than state of the art near-field gates require. We also present the laser systems required to for single-shot ablation loading and Doppler cooling of Be ions: a 70 mW 313 nm Dopper cooling laser, a 1064 nm ns-pulse laser for ablation of neutral beryllium atoms and a 235 nm laser for ionizing them. The detection system features an in vacuum imaging system on a cryogenic 3D translation stage that is able to scan and image the complete surface of an ion trap. The light can be detected on a photomultiplier tube or spatially resolved on an EMCCD camera. Our trap features a meander structure for driving gates and can trap up to 5 ions of which we, so far, did not lose any due to background gas collisions. This hints at a vacuum pressure of 1e-12 mbar or lower. The demonstrated system could hold a cryogenic QCCD chip with a few 10 qubits"}5{"work_id": 4404401018, "title": "PowerInfer: Fast Large Language Model Serving with a Consumer-grade GPU", "abstract": "This paper introduces PowerInfer, a high-speed Large Language Model (LLM) inference engine on a personal computer (PC) equipped with a single consumer-grade GPU. The key principle underlying the design of PowerInfer is exploiting the high locality inherent in LLM inference, characterized by a power-law distribution in neuron activation. This distribution indicates that a small subset of neurons, termed hot neurons, are consistently activated across inputs, while the majority, cold neurons, vary based on specific inputs. PowerInfer exploits such an insight to design a GPU-CPU hybrid inference engine: hot-activated neurons are preloaded onto the GPU for fast access, while cold-activated neurons are computed on the CPU, thus significantly reducing GPU memory demands and CPU-GPU data transfers. PowerInfer further integrates adaptive predictors and neuron-aware sparse operators, optimizing the efficiency of neuron activation and computational sparsity. The evaluation shows that PowerInfer significantly outperforms llama.cpp by up to 11.69\u00d7 while retaining model accuracy across various LLMs (including OPT-175B) on a single NVIDIA RTX 4090 GPU. For the OPT-30B model, PowerInfer achieves performance comparable to that of a high-end server-grade A100 GPU, reaching 82% of its token generation rate on a single consumer-grade RTX 4090 GPU."}6{"work_id": 7128511066, "title": "The impacts of globalization, institutional quality, and digital technology innovation on OECD countries sustainable digital economic growth", "abstract": "This study examines the effects of the digital economy components (digital technology innovation), institutional quality, and globalization on achieving sustainable digital economic growth within the Organization for Economic Cooperation and Development. Using panel data from 2002 to 2020 and applying the Generalized Method of Moments, the cointegration results confirm the presence of cointegration among all variables. The empirical findings reveal that digital technology innovation and institutional quality significantly enhance economic growth, with positive contributions of 0.014% and 0.011%, respectively. Meanwhile, globalization contributes by 0.533%. The study concludes that technological innovation, institutional quality, and globalization played a significant role in the country's economic expansion. The green total factor productivity shows a positive contribution. The Dumitrescu-Hurlin panel causality test results indicate two-way relationships among economic growth (GDP), labor, capital, institutional quality, and globalization. According to this analysis, selected nations can achieve the Sustainable Development Goals by implementing the necessary measures to achieve a balanced approach to sustainability across economics, society, and the environment. This can be accomplished through digital technology innovations, green technologies, and robust institutions. Integrating globalization and digitalization through increased investment in digital infrastructure would enhance access to innovative technologies and connectivity, while additionally improving efficiency."}7{"work_id": 4411368489, "title": "Making AI Less 'Thirsty'", "abstract": "Uncovering and addressing the secret water footprint of AI models"}8{"work_id": 2982169647, "title": "Quantum supremacy using a programmable superconducting processor", "abstract": "The promise of quantum computers is that certain computational tasks might be executed exponentially faster on a quantum processor than on a classical processor1. A fundamental challenge is to build a high-fidelity processor capable of running quantum algorithms in an exponentially large computational space. Here we report the use of a processor with programmable superconducting qubits2\u20137 to create quantum states on 53 qubits, corresponding to a computational state-space of dimension 253 (about 1016). Measurements from repeated experiments sample the resulting probability distribution, which we verify using classical simulations. Our Sycamore processor takes about 200 seconds to sample one instance of a quantum circuit a million times\u2014our benchmarks currently indicate that the equivalent task for a state-of-the-art classical supercomputer would take approximately 10,000 years. This dramatic increase in speed compared to all known classical algorithms is an experimental realization of quantum supremacy8\u201314 for this specific computational task, heralding a much-anticipated computing paradigm. Quantum supremacy is demonstrated using a programmable superconducting processor known as Sycamore, taking approximately 200 seconds to sample one instance of a quantum circuit a million times, which would take a state-of-the-art supercomputer around ten thousand years to compute."}9{"work_id": 2916516729, "title": "SLM-DB: Single-Level Key-Value Store with Persistent Memory", "abstract": "This paper investigates how to leverage emerging byteaddressable persistent memory (PM) to enhance the performance of key-value (KV) stores. We present a novel KV store, the Single-Level Merge DB (SLM-DB), which takes advantage of both the B+-tree index and the Log-Structured Merge Trees (LSM-tree) approach by making the best use of fast persistent memory. Our proposed SLM-DB achieves high read performance as well as high write performance with low write amplification and near-optimal read amplification. In SLM-DB, we exploit persistent memory to maintain a B+-tree index and adopt an LSM-tree approach to stage inserted KV pairs in a PM resident memory buffer. SLM-DB has a single-level organization of KV pairs on\\ndisks and performs selective compaction for the KV pairs, collecting garbage and keeping the KV pairs sorted sufficiently for range query operations. Our extensive experimental study demonstrates that, in our default setup, compared to LevelDB, SLM-DB provides 1.07 - 1.96 and 1.56 - 2.22 times higher read and write throughput, respectively, as well as comparable range query performance."}10{"work_id": 4401211704, "title": "Splitwise: Efficient Generative LLM Inference Using Phase Splitting", "abstract": "Generative large language model (LLM) applications are growing rapidly, leading to large-scale deployments of expensive and power-hungry GPUs. Our characterization of LLM inference shows that each inference request undergoes two phases: a compute-intensive prompt computation phase and a memory intensive token generation phase, each with distinct latency, throughput, memory, and power characteristics. Despite state-of-the-art batching and scheduling, the token generation phase underutilizes compute resources. Unlike prompt computation, token generation does not need the compute capability of the latest GPUs and can be run with lower power and cost. Based on these insights, we propose Splitwise, a model deployment and scheduling technique that splits the two phases of LLM inference requests on to separate machines. Splitwise enables phase-specific resource management using hardware that is well suited for each phase. Request state is transferred efficiently between machines using optimized network libraries on the fast back-plane interconnects available in today\u2019s GPU clusters. Using Splitwise, we design homogeneous and heterogeneous LLM inference clusters optimized for throughput, cost, and power Compared to current designs, Splitwise clusters achieve up to $1.4 \\times$ higher throughput at $\\mathbf{2 0 \\%}$ lower cost. Alternatively, they can deliver $2.35 \\times$ more throughput under the same power and cost budgets."}11{"work_id": 3010663313, "title": "FlatStore: An Efficient Log-Structured Key-Value Storage Engine for Persistent Memory", "abstract": "Emerging hardware like persistent memory (PM) and high-speed NICs are promising to build efficient key-value stores. However, we observe that the small-sized access pattern in key-value stores doesn't match with the persistence granularity in PMs, leaving the PM bandwidth underutilized. This paper proposes an efficient PM-based key-value storage engine named FlatStore. Specifically, it decouples the role of a KV store into a persistent log structure for efficient storage and a volatile index for fast indexing. Upon it, FlatStore further incorporates two techniques: 1) compacted log format to maximize the batching opportunity in the log; 2) pipelined horizontal batching to steal log entries from other cores when creating a batch, thus delivering low-latency and high-throughput performance. We implement FlatStore with the volatile index of both a hash table and Masstree. We deploy FlatStore on Optane DC Persistent Memory, and our experiments show that FlatStore achieves up to 35 Mops/s with a single server node, 2.5 - 6.3 times faster than existing systems."}12{"work_id": 4380874786, "title": "TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings", "abstract": "In response to innovations in machine learning (ML) models, production workloads changed radically and rapidly. TPU v4 is the fifth Google domain specific architecture (DSA) and its third supercomputer for such ML models. Optical circuit switches (OCSes) dynamically reconfigure its interconnect topology to improve scale, availability, utilization, modularity, deployment, security, power, and performance; users can pick a twisted 3D torus topology if desired. Much cheaper, lower power, and faster than Infiniband, OCSes and underlying optical components are <5% of system cost and <3% of system power. Each TPU v4 includes SparseCores, dataflow processors that accelerate models that rely on embeddings by 5x--7x yet use only 5% of die area and power. Deployed since 2020, TPU v4 outperforms TPU v3 by 2.1x and improves performance/Watt by 2.7x. The TPU v4 supercomputer is 4x larger at 4096 chips and thus nearly 10x faster overall, which along with OCS flexibility and availability allows a large language model to train at an average of ~60% of peak FLOPS/second. For similar sized systems, it is ~4.3x--4.5x faster than the Graphcore IPU Bow and is 1.2x--1.7x faster and uses 1.3x--1.9x less power than the Nvidia A100. TPU v4s inside the energy-optimized warehouse scale computers of Google Cloud use ~2--6x less energy and produce ~20x less CO2e than contemporary DSAs in typical on-premise data centers."}13{"work_id": 3138262672, "title": "SimuBoost: Scalable Parallelization of Functional System Simulation", "abstract": "F\u00fcr das Sammeln detaillierter Laufzeitinformationen, wie Speicherzugriffsmustern, wird in der Betriebssystem- und Sicherheitsforschung h\u00e4ufig auf die funktionale Systemsimulation zur\u00fcckgegriffen. Der Simulator f\u00fchrt dabei die zu untersuchende Arbeitslast in einer virtuellen Maschine (VM) aus, indem er schrittweise Instruktionen interpretiert oder derart \u00fcbersetzt, sodass diese auf dem Zustand der VM arbeiten. Dieser Prozess erm\u00f6glicht es, eine umfangreiche Instrumentierung durchzuf\u00fchren und so an Informationen zum Laufzeitverhalten zu gelangen, die auf einer physischen Maschine nicht zug\u00e4nglich sind. Obwohl die funktionale Systemsimulation als m\u00e4chtiges Werkzeug gilt, stellt die durch die Interpretation oder \u00dcbersetzung resultierende immense Ausf\u00fchrungsverlangsamung eine substanzielle Einschr\u00e4nkung des Verfahrens dar. Im Vergleich zu einer nativen Ausf\u00fchrung messen wir f\u00fcr QEMU eine 30-fache Verlangsamung, wobei die Aufzeichnung von Speicherzugriffen diesen Faktor verdoppelt. Mit Simulatoren, die umfangreichere Instrumentierungsm\u00f6glichkeiten mitbringen als QEMU, kann die Verlangsamung um eine Gr\u00f6\u00dfenordnung h\u00f6her ausfallen. Dies macht die funktionale Simulation f\u00fcr lang laufende, vernetzte oder interaktive Arbeitslasten uninteressant. Dar\u00fcber hinaus erzeugt die Verlangsamung ein unrealistisches Zeitverhalten, sobald Aktivit\u00e4ten au\u00dferhalb der VM (z. B. Ein-/Ausgabe) involviert sind. In dieser Arbeit stellen wir SimuBoost vor, eine Methode zur drastischen Beschleunigung funktionaler Systemsimulation. SimuBoost f\u00fchrt die zu untersuchende Arbeitslast zun\u00e4chst in einer schnellen hardwaregest\u00fctzten virtuellen Maschine aus. Dies erm\u00f6glicht volle Interaktivit\u00e4t mit Benutzern und Netzwerkger\u00e4ten. W\u00e4hrend der Ausf\u00fchrung erstellt SimuBoost periodisch Abbilder der VM (engl. Checkpoints). Diese dienen als Ausgangspunkt f\u00fcr eine parallele Simulation, bei der jedes Intervall unabh\u00e4ngig simuliert und analysiert wird. Eine heterogene deterministische Wiederholung (engl. heterogeneous deterministic Replay) garantiert, dass in dieser Phase die vorherige hardwaregest\u00fctzte Ausf\u00fchrung jedes Intervalls exakt reproduziert wird, einschlie\u00dflich Interaktionen und realistischem Zeitverhalten. Unser Prototyp ist in der Lage, die Laufzeit einer funktionalen Systemsimulation deutlich zu reduzieren. W\u00e4hrend mit herk\u00f6mmlichen Verfahren f\u00fcr die Simulation des Bauprozesses eines modernen Linux \u00fcber 5 Stunden ben\u00f6tigt werden, schlie\u00dft SimuBoost die Simulation in nur 15 Minuten ab. Dies sind lediglich 16% mehr Zeit, als der Bau in einer schnellen hardwaregest\u00fctzten VM in Anspruch nimmt. SimuBoost ist imstande, diese Geschwindigkeit auch bei voller Instrumentierung zur Aufzeichnung von Speicherzugriffen beizubehalten. Die vorliegende Arbeit ist das erste Projekt, welches das Konzept der Partitionierung und Parallelisierung der Ausf\u00fchrungszeit auf die interaktive Systemvirtualisierung in einer Weise anwendet, die eine sofortige parallele funktionale Simulation gestattet. Wir erg\u00e4nzen die praktische Umsetzung mit einem mathematischen Modell zur formalen Beschreibung der Beschleunigungseigenschaften. Dies erlaubt es, f\u00fcr ein gegebenes Szenario die voraussichtliche parallele Simulationszeit zu prognostizieren und gibt eine Orientierung zur Wahl der optimalen Intervalll\u00e4nge. Im Gegensatz zu bisherigen Arbeiten legt SimuBoost einen starken Fokus auf die Skalierbarkeit \u00fcber die Grenzen eines einzelnen physischen Systems hinaus. Ein zentraler Schl\u00fcssel hierzu ist der Einsatz moderner Checkpointing-Technologien. Im Rahmen dieser Arbeit pr\u00e4sentieren wir zwei neuartige Methoden zur effizienten und effektiven Kompression von periodischen Systemabbildern."}14{"work_id": 7129692983, "title": "Spatio-Temporal Coordinated Operation Strategy of Data Centers Considering Virtual Storage System via Two-Stage Distributionally Robust Optimization", "abstract": "Energy consumption in data centers is growing rapidly in recent years. To realize the economic and flexible operation of data centers, this paper proposes a spatio-temporal coordinated operation strategy of data centers considering virtual storage system (VSS) via two-stage distributionally robust optimization. First, based on the dynamic thermal network of buildings and differentiated demands of cloud users in data centers, a comprehensive operation model with VSS is proposed. Then, considering uncertainties of general load demands from data centers and outdoor temperatures, a fuzzy set is deduced, using an imprecise Dirichlet model derived from historical data. This ensures confidence level satisfaction for uncertain sets of general load and outdoor temperatures. In addition, the spatio-temporal coordinated operation strategy of data centers considering virtual storage system via two-stage distributionally robust optimization that flexibly adjusts conservativeness through uncertainty regulation parameters. Finally, a Dynamic Accuracy Column and Constraint Generation (DA-C&CG) algorithm is developed for solving the proposed strategy. The results indicate that the proposed strategy can effectively enhance operational economics, computing efficiency and resilience by leveraging flexibility in cooling and cloud user loads within suitable server temperatures. Meanwhile, the DA-C&CG algorithm exhibits excellent solution performance."}15{"work_id": 4294786246, "title": "Time-optimal holonomic quantum computation", "abstract": "A three-level system can be used in a \u039b-type configuration in order to construct auniversal set of non-adiabatic quantum gates through the use of non-Abelian non-adiabatic geometrical phases. Such construction allows for high-speed operation times which diminish the effects of decoherence. This might be, however, accompanied by a breakdown of the validity of the rotating wave approximation (RWA) due to the comparable timescale between the counter-rotating terms and the pulse length, which greatly affects the dynamics. Here we investigate the trade-off between dissipative effects and the RWA validity, obtaining the optimal regime for the operation of the holonomic quantum gates."}16{"work_id": 4394944658, "title": "Orion: Interference-aware, Fine-grained GPU Sharing for ML Applications", "abstract": "GPUs are critical for maximizing the throughput-per-Watt of deep neural network (DNN) applications. However, DNN applications often underutilize GPUs, even when using large batch sizes and eliminating input data processing or communication stalls. DNN workloads consist of data-dependent operators, with different compute and memory requirements. While an operator may saturate GPU compute units or memory bandwidth, it often leaves other GPU resources idle. Despite the prevalence of GPU sharing techniques, current approaches are not sufficiently fine-grained or interference-aware to maximize GPU utilization while minimizing interference at the granularity of 10s of \u03bcs. We propose Orion, a system that transparently intercepts GPU kernel launches from multiple clients sharing a GPU. Orion schedules work on the GPU at the granularity of individual operators and minimizes interference by taking into account each operator's compute and memory requirements. We integrate Orion in PyTorch and demonstrate its benefits in various DNN workload collocation use cases. Orion significantly improves tail latency compared to state-of-the-art baselines for a high-priority inference job while collocating best-effort inference jobs to increase per-GPU request throughput by up to 7.3\u00d7, or while collocating DNN training, saving up to 1.49\u00d7 in training costs compared to dedicated GPU allocation."}17{"work_id": 4295682935, "title": "CodaLab Competitions An Open Source Platform to Organize Scientific Challenges", "abstract": "CodaLab Competitions is an open source web platform designed to help data scientists and research teams to crowd-source the resolution of machine learning problems through the organization of competitions, also called challenges or contests. CodaLab Competitions provides useful features such as multiple phases, results and code submissions, multi-score leaderboards, and jobs running inside Docker containers. The platform is very flexible and can handle large scale experiments, by allowing organizers to upload large datasets and provide their own CPU or GPU compute workers."}18{"work_id": 4407091026, "title": "Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): a randomised, controlled, parallel-group, non-inferiority, single-blinded, screening accuracy study", "abstract": "BACKGROUND: Emerging evidence suggests that artificial intelligence (AI) can increase cancer detection in mammography screening while reducing screen-reading workload, but further understanding of the clinical impact is needed. METHODS: In this randomised, controlled, parallel-group, non-inferiority, single-blinded, screening-accuracy study, done within the Swedish national screening programme, women recruited at four screening sites in southwest Sweden (Malm\u00f6, Lund, Landskrona, and Trelleborg) who were eligible for mammography screening were randomly allocated (1:1) to AI-supported screening or standard double reading. The AI system (Transpara version 1.7.0 ScreenPoint Medical, Nijmegen, Netherlands) was used to triage screening examinations to single or double reading and as detection support highlighting suspicious findings. This is a protocol-defined analysis of the secondary outcome measures of recall, cancer detection, false-positive rates, positive predictive value of recall, type and stage of cancer detected, and screen-reading workload. This trial is registered at ClinicalTrials.gov, NCT04838756 and is closed to accrual. FINDINGS: Between April 12, 2021, and Dec 7, 2022, 105 934 women were randomly assigned to the intervention or control group. 19 women were excluded from the analysis. The median age was 53\u00b77 years (IQR 46\u00b75-63\u00b72). AI-supported screening among 53 043 participants resulted in 338 detected cancers and 1110 recalls. Standard screening among 52 872 participants resulted in 262 detected cancers and 1027 recalls. Cancer-detection rates were 6\u00b74 per 1000 (95% CI 5\u00b77-7\u00b71) screened participants in the intervention group and 5\u00b70 per 1000 (4\u00b74-5\u00b76) in the control group, a ratio of 1\u00b729 (95% CI 1\u00b709-1\u00b751; p=0\u00b70021). AI-supported screening resulted in an increased detection of invasive cancers (270 vs 217, a proportion ratio of 1\u00b724 [95% CI 1\u00b704-1\u00b748]), wich were mainly small lymph-node negative cancers (58 more T1, 46 more lymph-node negative, and 21 more non-luminal A). AI-supported screening also resulted in an increased detection of in situ cancers (68 vs 45, a proportion ratio of 1\u00b751 [1\u00b703-2\u00b719]), with about half of the increased detection being high-grade in situ cancer (12 more nuclear grade III, and no increase in nuclear grade I). The recall and false-positive rate were not significantly higher in the intervention group (a ratio of 1\u00b708 [95% CI 0\u00b799-1\u00b717; p=0\u00b7084] and 1\u00b701 [0\u00b791-1\u00b711; p=0\u00b792], respectively). The positive predictive value of recall was significantly higher in the intervention group compared with the control group, with a ratio of 1\u00b719 (95% CI 1\u00b704-1\u00b737; p=0\u00b7012). There were 61 248 screen readings in the intervention group and 109 692 in the control group, resulting in a 44\u00b72% reduction in the screen-reading workload. INTERPRETATION: The findings suggest that AI contributes to the early detection of clinically relevant breast cancer and reduces screen-reading workload without increasing false positives. FUNDING: Swedish Cancer Society, Confederation of Regional Cancer Centres, and Swedish governmental funding for clinical research."}19{"work_id": 3037032032, "title": "Energy and Policy Considerations for Modern Deep Learning Research", "abstract": "The field of artificial intelligence has experienced a dramatic methodological shift towards large neural networks trained on plentiful data. This shift has been fueled by recent advances in hardware and techniques enabling remarkable levels of computation, resulting in impressive advances in AI across many applications. However, the massive computation required to obtain these exciting results is costly both financially, due to the price of specialized hardware and electricity or cloud compute time, and to the environment, as a result of non-renewable energy used to fuel modern tensor processing hardware. In a paper published this year at ACL, we brought this issue to the attention of NLP researchers by quantifying the approximate financial and environmental costs of training and tuning neural network models for NLP (Strubell, Ganesh, and McCallum 2019). In this extended abstract, we briefly summarize our findings in NLP, incorporating updated estimates and broader information from recent related publications, and provide actionable recommendations to reduce costs and improve equity in the machine learning and artificial intelligence community."}20{"work_id": 3107463944, "title": "Quantum computational advantage using photons", "abstract": "A light approach to quantum advantage Quantum computational advantage or supremacy is a long-anticipated milestone toward practical quantum computers. Recent work claimed to have reached this point, but subsequent work managed to speed up the classical simulation and pointed toward a sample size\u2013dependent loophole. Quantum computational advantage, rather than being a one-shot experimental proof, will be the result of a long-term competition between quantum devices and classical simulation. Zhong et al. sent 50 indistinguishable single-mode squeezed states into a 100-mode ultralow-loss interferometer and sampled the output using 100 high-efficiency single-photon detectors. By obtaining up to 76-photon coincidence, yielding a state space dimension of about 10 30 , they measured a sampling rate that is about 10 14 -fold faster than using state-of-the-art classical simulation strategies and supercomputers. Science , this issue p. 1460"}21{"work_id": 1868882999, "title": "Formal verification techniques using quantum process calculus", "abstract": "Quantum communication is a rapidly growing area of research and development. While the successful construction of a large-scale quantum computer may be some years away, there are already commercial implementations of secure communication using quantum cryptography. The application of formal methods to classical communication and cryptographic systems has been very successful, and is now widely used in industry by organisations such as Intel, Microsoft and NASA. There is reason to believe that similar benefits can be expected for the verification of quantum systems. In this thesis, we focus on the use of process calculus, specifically Communicating Quantum Processes (CQP), for the analysis of quantum protocols. Congruence relations are an important aspect of process calculus, since they provide the foundation for equational reasoning. Previous work on congruence relations for quantum processes excluded the classical information arising from measurements, and was therefore unable to analyse many of the interesting known quantum communication protocols. Developing a congruence relation for general quantum processes is difficult because of the interaction between measurement, entanglement and parallel composition. We define a labelled transition relation for CQP in order to describe external interactions. Based on this semantics, we define a notion of observational equivalence for CQP processes, namely probabilistic branching bisimilarity. We find that this relation is not preserved by parallel composition, however we are able to gain a deeper understanding of the link between probabilistic branching and measurement. Based on this newfound understanding, we present a novel semantics for quantum processes, combining mixed quantum states with probabilistic branching. With respect to this new semantic model, we define full probabilistic branching bisimilarity and prove that it is a congruence. We use this congruence relation to discuss an axiomatic approach to the verification of quantum processes. The quantum teleportation protocol is used as a primary example throughout, and we prove that it is congruent to a quantum channel. We define a translation from CQP to the Quantum Model Checker (QMC) in order to provide automated verification techniques using CQP specifications. We prove that this translation preserves the semantics of CQP processes, thereby enabling a multifaceted approach to formal verification by enhancing the manual techniques of process calculus with the benefits of model checking."}22{"work_id": 2931743911, "title": "Machine Learning at Facebook: Understanding Inference at the Edge", "abstract": "At Facebook, machine learning provides a wide range of capabilities that drive many aspects of user experience including ranking posts, content understanding, object detection and tracking for augmented and virtual reality, speech and text translations. While machine learning models are currently trained on customized data-center infrastructure, Facebook is working to bring machine learning inference to the edge. By doing so, user experience is improved with reduced latency (inference time) and becomes less dependent on network connectivity. Furthermore, this also enables many more applications of deep learning with important features only made available at the edge. This paper takes a data-driven approach to present the opportunities and design challenges faced by Facebook in order to enable machine learning inference locally on smart phones and other edge platforms."}23{"work_id": 4210305137, "title": "Photonic matrix multiplication lights up photonic accelerator and beyond", "abstract": "Matrix computation, as a fundamental building block of information processing in science and technology, contributes most of the computational overheads in modern signal processing and artificial intelligence algorithms. Photonic accelerators are designed to accelerate specific categories of computing in the optical domain, especially matrix multiplication, to address the growing demand for computing resources and capacity. Photonic matrix multiplication has much potential to expand the domain of telecommunication, and artificial intelligence benefiting from its superior performance. Recent research in photonic matrix multiplication has flourished and may provide opportunities to develop applications that are unachievable at present by conventional electronic processors. In this review, we first introduce the methods of photonic matrix multiplication, mainly including the plane light conversion method, Mach-Zehnder interferometer method and wavelength division multiplexing method. We also summarize the developmental milestones of photonic matrix multiplication and the related applications. Then, we review their detailed advances in applications to optical signal processing and artificial neural networks in recent years. Finally, we comment on the challenges and perspectives of photonic matrix multiplication and photonic acceleration."}24{"work_id": 4393145114, "title": "AI and Memory Wall", "abstract": "The availability of unprecedented unsupervised training data, along with neural scaling laws, has resulted in an unprecedented surge in model size and compute requirements for serving/training LLMs. However, the main performance bottleneck is increasingly shifting to memory bandwidth. Over the past 20 years, peak server hardware FLOPS has been scaling at 3.0\u00d7/2yrs, outpacing the growth of DRAM and interconnect bandwidth, which have only scaled at 1.6 and 1.4 times every 2 years, respectively. This disparity has made memory, rather than compute, the primary bottleneck in AI applications, particularly in serving. Here, we analyze encoder and decoder Transformer models and show how memory bandwidth can become the dominant bottleneck for decoder models. We argue for a redesign in model architecture, training, and deployment strategies to overcome this memory limitation."}25{"work_id": 3173899054, "title": "Strong Quantum Computational Advantage Using a Superconducting Quantum Processor", "abstract": "Scaling up to a large number of qubits with high-precision control is essential in the demonstrations of quantum computational advantage to exponentially outpace the classical hardware and algorithmic improvements. Here, we develop a two-dimensional programmable superconducting quantum processor, Zuchongzhi, which is composed of 66 functional qubits in a tunable coupling architecture. To characterize the performance of the whole system, we perform random quantum circuits sampling for benchmarking, up to a system size of 56 qubits and 20 cycles. The computational cost of the classical simulation of this task is estimated to be 2-3 orders of magnitude higher than the previous work on 53-qubit Sycamore processor [Nature 574, 505 (2019)NATUAS0028-083610.1038/s41586-019-1666-5. We estimate that the sampling task finished by Zuchongzhi in about 1.2 h will take the most powerful supercomputer at least 8 yr. Our work establishes an unambiguous quantum computational advantage that is infeasible for classical computation in a reasonable amount of time. The high-precision and programmable quantum computing platform opens a new door to explore novel many-body phenomena and implement complex quantum algorithms."}26{"work_id": 3092732687, "title": "PENGEMBANGAN SISTEM INFORMASI PENERIMAANMAHASISWA BARU DENGAN METODE BUSINESSPROCESS AUTOMATION (STUDI KASUS: INSTITUTTEKNOLOGI KALIMANTAN)", "abstract": "Institut Teknologi Kalimantan (ITK) merupakan salah satu perguruan tinggi negeri baru di Indonesia. Salah satu proses bisnis yang ada di ITK yaitu proses\r\npenerimaan mahasiswa baru, ITK sudah melakukan proses penerimaan mahasiswa baru sejak tahun 2012 sebanyak 100 mahasiswa, dan pada tahun 2018 ITK menerima kurang lebih 1000 mahasiswa. Hal ini menunjukkan adanya peningkatan penerimaan mahasiswa baru yang cukup signifikan. Pada proses penerimaan mahasiswa baru di ITK belum ada sistem informasi yangdapat memanajemen data\r\nmahasiswa baru tersebut dan saat ini terdapat proses pengumpulan dan verifikasi\r\nberkas yang masih dilakukan menggunakan email dimana terdapat beberapa kendala dalam penggunaan email tersebut. Oleh karena itu, dalam penelitian ini\r\ndilakukan pengembangkan sistem informasi penerimaan mahasiswa baru berbasis web, yang dapat diakses oleh mahasiswa baru, verifikatur, dan admin selama proses\r\npendaftaran ulang mahasiswa baru. Metode yang digunakan dalam penelitian ini adalah Business Proses Automation menggunakan tools ProcessMaker. ProcessMaker merupakan sebuah aplikasi open source yang berguna untuk memanajemen alur kerja Business Process Management (BPM) dan solusi perangkat lunak berbasis web. Penelitian ini menghasilkan sistem informasi penerimaan mahasiswa baru berbasis web, yang dapat memanajemen datamahasiswa baru. Terdapat 14 proses pendaftaran ulang mahasiswa baru dan rekapdata yang dibedakan untuk masing-masing program studi di ITK, kemudian ada 1 proses untuk menjalankan fungsi impor data users menggunakan file .csv. Sistem\r\ninformasi ini sudah mulai digunakan ITK dalam proses penerimaan mahasiswa baru\r\njalur Seleksi Nasional Masuk Perguruan Tinggi Negeri (SNMPTN) Tahun 2019."}27{"work_id": 25727798, "title": "The Lustre Storage Architecture", "abstract": "This lengthy document often referred to as the \"Lustre Book\", contains a detailed outline of Lustre file system architecture, as it was created between 2001 and 2005, in accordance with the requirements from various users. Now, in 2019, most features have been implemented, but some only recently, and some along different lines of thought."}28{"work_id": 2972087877, "title": "Demystifying Parallel and Distributed Deep Learning", "abstract": "Deep Neural Networks (DNNs) are becoming an important tool in modern computing applications. Accelerating their training is a major challenge and techniques range from distributed algorithms to low-level circuit design. In this survey, we describe the problem from a theoretical perspective, followed by approaches for its parallelization. We present trends in DNN architectures and the resulting implications on parallelization strategies. We then review and model the different types of concurrency in DNNs: from the single operator, through parallelism in network inference and training, to distributed deep learning. We discuss asynchronous stochastic optimization, distributed system architectures, communication schemes, and neural architecture search. Based on those approaches, we extrapolate potential directions for parallelism in deep learning."}29{"work_id": 4407158326, "title": "Distributed quantum computing across an optical network link", "abstract": "Abstract Distributed quantum computing (DQC) combines the computing power of multiple networked quantum processing modules, ideally enabling the execution of large quantum circuits without compromising performance or qubit connectivity 1,2 . Photonic networks are well suited as a versatile and reconfigurable interconnect layer for DQC; remote entanglement shared between matter qubits across the network enables all-to-all logical connectivity through quantum gate teleportation (QGT) 3,4 . For a scalable DQC architecture, the QGT implementation must be deterministic and repeatable; until now, no demonstration has satisfied these requirements. Here we experimentally demonstrate the distribution of quantum computations between two photonically interconnected trapped-ion modules. The modules, separated by about two metres, each contain dedicated network and circuit qubits. By using heralded remote entanglement between the network qubits, we deterministically teleport a controlled-Z (CZ) gate between two circuit qubits in separate modules, achieving 86% fidelity. We then execute Grover\u2019s search algorithm 5 \u2014to our knowledge, the first implementation of a distributed quantum algorithm comprising several non-local two-qubit gates\u2014and measure a 71% success rate. Furthermore, we implement distributed iSWAP and SWAP circuits, compiled with two and three instances of QGT, respectively, demonstrating the ability to distribute arbitrary two-qubit operations 6 . As photons can be interfaced with a variety of systems, the versatile DQC architecture demonstrated here provides a viable pathway towards large-scale quantum computing for a range of physical platforms."}30{"work_id": 3152809665, "title": "Nightcore: efficient and scalable serverless computing for latency-sensitive, interactive microservices", "abstract": "The microservice architecture is a popular software engineering approach for building flexible, large-scale online services. Serverless functions, or function as a service (FaaS), provide a simple programming model of stateless functions which are a natural substrate for implementing the stateless RPC handlers of microservices, as an alternative to containerized RPC servers. However, current serverless platforms have millisecond-scale runtime overheads, making them unable to meet the strict sub-millisecond latency targets required by existing interactive microservices."}31{"work_id": 4399913870, "title": "A review of green artificial intelligence: Towards a more sustainable future", "abstract": "Green artificial intelligence (AI) is more environmentally friendly and inclusive than conventional AI, as it not only produces accurate results without increasing the computational cost but also ensures that any researcher with a laptop can perform high-quality research without the need for costly cloud servers. This paper discusses green AI as a pivotal approach to enhancing the environmental sustainability of AI systems. Described are AI solutions for eco-friendly practices in other fields (green-by AI), strategies for designing energy-efficient machine learning (ML) algorithms and models (green-in AI), and tools for accurately measuring and optimizing energy consumption. Also examined are the role of regulations in promoting green AI and future directions for sustainable ML. Underscored is the importance of aligning AI practices with environmental considerations, fostering a more eco-conscious and energy-efficient future for AI systems."}32{"work_id": 4409363528, "title": "Intelligent nanophotonics: when machine learning sheds light", "abstract": "Abstract The synergistic development of nanophotonics and machine learning has inspired tremendous innovations in both fields in the past decade. In diverse photonics research, deep-learning methods using artificial neural networks become the key game changer that greatly facilitates rapid nanophotonics design and the versatile processing of optical information. Moreover, optical computing platforms that perform calculations through light propagation are receiving tremendous interest as next-generation machine-learning hardware with advantages in computing speed, energy efficiency, and parallelism. This review summarizes the current state-of-the-art nanophotonic devices enabled by machine learning and analyzes the longstanding challenges that must be overcome to make an impact on technology. We also discuss the opportunities of intelligent photonics in applications such as computational imaging/sensing and machine vision. The intersection of nanophotonics with deep learning holds tremendous implications for transformative technologies ranging from internet of things to smart health. Lastly, we provide our perspective on the pressing challenges in intelligent photonics that must be tackled to advance this field to the next level and the vast opportunities for multidisciplinary collaboration."}33{"work_id": 4386978266, "title": "Accelerated Systems for Portable DNA Sequencing", "abstract": "The MinION is a revolutionary handheld DNA sequencer that is inexpensive, portable, and can perform real-time sequencing. MinION is increasingly used in Precision Medicine applications. However, the MinION lacks portable compute power. This thesis introduces two clinical applications of the MinION and identifies and solves performance bottlenecks through hardware and software solutions to enable portable microbial diagnostics. Finally, we discuss how our accelerated solutions will fit on a laptop with a GPU. More than 99% of DNA fragments in a typical human sample are non-target (human), which may be skipped in real-time using the MinION\u2019s Read Until feature. We analyze the performance of the Read Until pipeline in detecting target microbial species for targeted viral pathogen detection and microbiome abundance estimation. We find new sources of performance bottlenecks (basecaller in the classification of a fragment) that are not addressed by past genomics accelerators. While SquiggleFilter and DTWax are our solutions for targeted viral pathogen detection, RawMap is for untargeted microbiome analysis. We also discuss accelerating the bottleneck step in the DNA mapping software (Minimap2) used in all of MinION\u2019s sequencing workflows. For targeted virus detection, we discuss SquiggleFilter which is a portable and programmable hardware-software solution that directly analyzes MinION\u2019s raw squiggles and filters everything except target viral DNA fragments. SquiggleFilter avoids the expensive basecalling step and uses hardware-accelerated subsequence Dynamic Time Warping (sDTW). We show that our 14.3W 13.25mm2 accelerator has 274\u00d7 greater throughput and 3481\u00d7 lower latency than existing GPU-based solutions while consuming half the power. DTWax overcomes the on-chip memory limitations of SquiggleFilter by optimizing its high-accuracy sDTW algorithm for GPUs resulting in a \u223c1.92X sequencing speedup and \u223c3.64X compute speedup: costup. For the untargeted classification and analysis of microbiome, we discuss RawMap which is a machine-learning-based smart and efficient solution that reduces sequencing time and cost by \u223c24% and computing cost by \u223c22%. We also discuss how RawMap may be used as an alternative to the RT-PCR test for viral load quantification of SARS-CoV-2. Minimap2 is a software used in all MinION workflows. minimap2-accelerated (mm2-ax) is a heterogeneous design for sequence mapping where minimap2\u2019s bottleneck step is sped up on the GPU with bit-exact output. mm2-ax on an NVIDIA A100 GPU improves the bottleneck step (chaining) with 4.07 - 1.93X speedup: costup over a SIMD baseline. Finally, we envision a portable solution to microbial diagnostics with a laptop connected to the MinION. DTWax can perform targeted virus detection on the GPU and RawMap does untargeted microbiome classification on the CPU. Post-sequencing tasks like basecalling, alignment (using mm2-ax) and variant calling use the GPU."}34{"work_id": 2950805899, "title": "A quantum engineer's guide to superconducting qubits", "abstract": "The aim of this review is to provide quantum engineers with an introductory guide to the central concepts and challenges in the rapidly accelerating field of superconducting quantum circuits. Over the past twenty years, the field has matured from a predominantly basic research endeavor to a one that increasingly explores the engineering of larger-scale superconducting quantum systems. Here, we review several foundational elements\u2014qubit design, noise properties, qubit control, and readout techniques\u2014developed during this period, bridging fundamental concepts in circuit quantum electrodynamics and contemporary, state-of-the-art applications in gate-model quantum computation."}35{"work_id": 2273211146, "title": "The multiprocessor real-time scheduling of general task systems", "abstract": "The recent emergence of multicore and related technologies in many commercial systems has increased the prevalence of multiprocessor architectures. Contemporaneously, real-time applications have become more complex and sophisticated in their behavior and interaction. Inevitably, these complex real-time applications will be deployed upon these multiprocessor platforms and require temporal analysis techniques to verify their correctness. However, most prior research in multiprocessor real-time scheduling has addressed the temporal analysis only of Liu and Layland task systems. The goal of this dissertation is to extend real-time scheduling theory for multiprocessor systems by developing temporal analysis techniques for more general task models such as the sporadic task model, the generalized multiframe task model, and the recurring real-time task model. The thesis of this dissertation is: Optimal online multiprocessor real-time scheduling algorithms for sporadic and more general task systems are impossible; however, efficient, online scheduling algorithms and associated feasibility and schedulability tests, with provably bounded deviation from any optimal test, exist. To support our thesis, this dissertation develops feasibility and schedulability tests for various multiprocessor scheduling paradigms. We consider three classes of multiprocessor scheduling based on whether a real-time job may migrate between processors: full-migration, restricted-migration, and partitioned. For all general task systems, we obtain feasibility tests for arbitrary real-time instances under the full-and restricted-migration paradigms. Despite the existence of tests for feasibility, we show that optimal online scheduling of sporadic and more general systems is impossible. Therefore, we focus on scheduling algorithms that have constant-factor approximation ratios in terms of an analysis technique known as resource augmentation. We develop schedulability tests for scheduling algorithms, earliest-deadline-first (edf) and deadline-monotonic (dm), under full-migration and partitioned scheduling paradigms. Feasibility and schedulability tests presented in this dissertation use the workload metrics of demand-based load and maximum job density and have provably bounded deviation from optimal in terms of resource augmentation. We show the demand-based load and maximum job density metrics may be exactly computed in pseudo-polynomial time for general task systems and approximated in polynomial time for sporadic task systems."}36{"work_id": 4385215131, "title": "QLC Programmable 3D Ferroelectric NAND Flash Memory by Memory Window Expansion using Cell Stack Engineering", "abstract": "3D ferroelectric NAND (Fe-NAND) Quad-level cell (QLC) operation has been demonstrated for the first time to our knowledge, using the 3D CTN NAND test vehicle for mass production. The 3D Fe-NAND is optimized by engineering the cell stack layers, enlarging a program/erase (PE) window up to 10.5 V. QLC operation is successfully verified with the minimum gap margin of 0.24 V. Endurance and data retention characteristics are also reported."}37{"work_id": 3174122836, "title": "Quantum Computing 40 Years Later", "abstract": "This chapter sketches the foundations of the subject, discussing salient features of quantum information, formulates a mathematical model of quantum computation and highlights some implications of the model. It reviews two particularly promising applications of quantum computing foreseen by Feynman, simulating the dynamics of complex quantum systems, and computing their static properties. The chapter explains the concept of quantum error correction, the basis of belief that quantum computers can be scaled up to large systems that solve very hard problems. Shor&s;s discovery, and its obvious implications for cryptanalysis, caused interest in quantum computing to skyrocket. Because of Shor&s;s algorithm, the public key cryptographic protocols the people use to protect privacy when they communicate over the Internet will become vulnerable to attacks by quantum computers in the future. Classical computers are especially bad at simulating quantum dynamics \u2013 that is, predicting how a highly entangled quantum state will change with time."}38{"work_id": 4407592994, "title": "Green Federated Learning: A New Era of Green Aware AI", "abstract": "The development of AI applications, especially in large-scale wireless networks, is growing exponentially, alongside the size and complexity of the architectures used. Particularly, machine learning is acknowledged as one of today\u2019s most energy-intensive computational applications, posing a significant challenge to the environmental sustainability of next-generation intelligent systems. Achieving environmental sustainability entails ensuring that every AI algorithm is designed with sustainability in mind, integrating green considerations from the architectural phase onwards. Recently, Federated Learning (FL), with its distributed nature, presents new opportunities to address this need. Hence, it is imperative to elucidate the potential and challenges stemming from recent FL advancements and their implications for sustainability. Moreover, it is crucial to furnish researchers, stakeholders, and interested parties with a roadmap to navigate and understand existing efforts and gaps in green-aware AI algorithms. This survey primarily aims to achieve this objective by identifying and analyzing over a hundred FL works and assessing their contributions to green-aware artificial intelligence for sustainable environments, with a specific focus on IoT research. It delves into current issues in green federated learning from an energy-efficient standpoint, discussing potential challenges and future prospects for green IoT application research."}39{"work_id": 4406650295, "title": "AWQ: Activation-aware Weight Quantization for On-Device LLM Compression and Acceleration", "abstract": "Large language models (LLMs) have transformed numerous AI applications. On-device LLM is becoming increasingly important: running LLMs locally on edge devices can reduce cloud computing costs and protect users' privacy. However, the astronomical model size and the limited hardware resources pose significant deployment challenges. To solve these issues, we propose Activation-aware Weight Quantization (AWQ) and TinyChat, an algorithm-system full-stack solution for efficient on-device LLM deployment. AWQ is a novel quantization method that identifies and protects salient weights based on activation distribution, significantly reducing model size while preserving performance. TinyChat, an optimized inference framework, translates AWQ's theoretical memory savings into practical speedups through techniques such as on-the-fly dequantization, SIMD-aware weight packing, and kernel fusion. Together, they enable 4x model size reduction and 3-4x acceleration across various edge platforms, from high-end desktop GPUs to resource-constrained IoT devices. This solution democratizes on-device LLM deployment, offering privacy-preserving, low-latency AI capabilities across a wide range of applications."}40{"work_id": 4213212652, "title": "Noisy intermediate-scale quantum algorithms", "abstract": "Noisy quantum computers can in principle perform reliable quantum computations, but truly scalable systems require noise levels lower than are presently achieved. Still, moderate-complexity computations can be performed. This review discusses what is possible in this ``noisy intermediate scale'' quantum (NISQ) era. Topic areas include the simulation of many-body physics and chemistry, combinatorial optimization, and machine learning. It is evident that the NISQ era has produced new paradigms for programming that will be built upon as quantum computers are further perfected."}41{"work_id": 2975712713, "title": "A generic communication scheduler for distributed DNN training acceleration", "abstract": "We present ByteScheduler, a generic communication scheduler for distributed DNN training acceleration. ByteScheduler is based on our principled analysis that partitioning and rearranging the tensor transmissions can result in optimal results in theory and good performance in real-world even with scheduling overhead. To make ByteScheduler work generally for various DNN training frameworks, we introduce a unified abstraction and a Dependency Proxy mechanism to enable communication scheduling without breaking the original dependencies in framework engines. We further introduce a Bayesian Optimization approach to auto-tune tensor partition size and other parameters for different training models under various networking conditions. ByteScheduler now supports TensorFlow, PyTorch, and MXNet without modifying their source code, and works well with both Parameter Server (PS) and all-reduce architectures for gradient synchronization, using either TCP or RDMA. Our experiments show that ByteScheduler accelerates training with all experimented system configurations and DNN models, by up to 196% (or 2.96X of original speed)."}42{"work_id": 2930508541, "title": "Cloud Programming Simplified: A Berkeley View on Serverless Computing", "abstract": "Serverless cloud computing handles virtually all the system administration operations needed to make it easier for programmers to use the cloud. It provides an interface that greatly simplifies cloud programming, and represents an evolution that parallels the transition from assembly language to high-level programming languages. This paper gives a quick history of cloud computing, including an accounting of the predictions of the 2009 Berkeley View of Cloud Computing paper, explains the motivation for serverless computing, describes applications that stretch the current limits of serverless, and then lists obstacles and research opportunities required for serverless computing to fulfill its full potential. Just as the 2009 paper identified challenges for the cloud and predicted they would be addressed and that cloud use would accelerate, we predict these issues are solvable and that serverless computing will grow to dominate the future of cloud computing."}43{"work_id": 3016932526, "title": "Q uantum ESPRESSO toward the exascale", "abstract": "Quantum ESPRESSO is an open-source distribution of computer codes for quantum-mechanical materials modeling, based on density-functional theory, pseudopotentials, and plane waves, and renowned for its performance on a wide range of hardware architectures, from laptops to massively parallel computers, as well as for the breadth of its applications. In this paper, we present a motivation and brief review of the ongoing effort to port Quantum ESPRESSO onto heterogeneous architectures based on hardware accelerators, which will overcome the energy constraints that are currently hindering the way toward exascale computing."}44{"work_id": 4402694963, "title": "Optical neural networks: progress and challenges", "abstract": "Artificial intelligence has prevailed in all trades and professions due to the assistance of big data resources, advanced algorithms, and high-performance electronic hardware. However, conventional computing hardware is inefficient at implementing complex tasks, in large part because the memory and processor in its computing architecture are separated, performing insufficiently in computing speed and energy consumption. In recent years, optical neural networks (ONNs) have made a range of research progress in optical computing due to advantages such as sub-nanosecond latency, low heat dissipation, and high parallelism. ONNs are in prospect to provide support regarding computing speed and energy consumption for the further development of artificial intelligence with a novel computing paradigm. Herein, we first introduce the design method and principle of ONNs based on various optical elements. Then, we successively review the non-integrated ONNs consisting of volume optical components and the integrated ONNs composed of on-chip components. Finally, we summarize and discuss the computational density, nonlinearity, scalability, and practical applications of ONNs, and comment on the challenges and perspectives of the ONNs in the future development trends."}45{"work_id": 4287393709, "title": "Uncertainty-Aware Resource Allocation for Multi-Path Programs with In-Kernel Predictions", "abstract": "Predictable timing on multicore systems requires careful management of shared resources such as the last-level cache and memory bandwidth. This paper presents MPORA, an uncertainty-aware dynamic resource allocation framework for multi-path, input-dependent real-time tasks on multicore platforms. MPORA models each job as a discrete-time dynamical system that captures execution dynamics and resource-dependent performance indicators. At runtime, MPORA monitors job execution states and predicts short-term instruction rates and remaining execution times under candidate allocations using predictive models trained offline. It then solves a receding-horizon optimization problem to compute resource allocations that maximize system-wide progress while meeting job deadlines. To address prediction uncertainty, MPORA integrates weighted conformal prediction into the optimization formulation, enabling uncertainty-aware deadline constraints. We implement MPORA as a Linux kernel module with microsecond-scale inference overhead. Experimental results on SPEC CPU benchmarks show that MPORA delivers accurate predictions under unseen inputs and distribution shifts with low overhead, while improving schedulability and response times over existing methods."}46{"work_id": 4321610982, "title": "Suppressing quantum errors by scaling a surface code logical qubit", "abstract": "Abstract Practical quantum computing will require error rates well below those achievable with physical qubits. Quantum error correction 1,2 offers a path to algorithmically relevant error rates by encoding logical qubits within many physical qubits, for which increasing the number of physical qubits enhances protection against physical errors. However, introducing more qubits also increases the number of error sources, so the density of errors must be sufficiently low for logical performance to improve with increasing code size. Here we report the measurement of logical qubit performance scaling across several code sizes, and demonstrate that our system of superconducting qubits has sufficient performance to overcome the additional errors from increasing qubit number. We find that our distance-5 surface code logical qubit modestly outperforms an ensemble of distance-3 logical qubits on average, in terms of both logical error probability over 25 cycles and logical error per cycle ((2.914 \u00b1 0.016)% compared to (3.028 \u00b1 0.023)%). To investigate damaging, low-probability error sources, we run a distance-25 repetition code and observe a 1.7 \u00d7 10 \u22126 logical error per cycle floor set by a single high-energy event (1.6 \u00d7 10 \u22127 excluding this event). We accurately model our experiment, extracting error budgets that highlight the biggest challenges for future systems. These results mark an experimental demonstration in which quantum error correction begins to improve performance with increasing qubit number, illuminating the path to reaching the logical error rates required for computation."}47{"work_id": 3156891177, "title": "Carbon Emissions and Large Neural Network Training", "abstract": "The computation demand for machine learning (ML) has grown rapidly recently, which comes with a number of costs. Estimating the energy cost helps measure its environmental impact and finding greener strategies, yet it is challenging without detailed information. We calculate the energy use and carbon footprint of several recent large models-T5, Meena, GShard, Switch Transformer, and GPT-3-and refine earlier estimates for the neural architecture search that found Evolved Transformer. We highlight the following opportunities to improve energy efficiency and CO2 equivalent emissions (CO2e): Large but sparsely activated DNNs can consume <1/10th the energy of large, dense DNNs without sacrificing accuracy despite using as many or even more parameters. Geographic location matters for ML workload scheduling since the fraction of carbon-free energy and resulting CO2e vary ~5X-10X, even within the same country and the same organization. We are now optimizing where and when large models are trained. Specific datacenter infrastructure matters, as Cloud datacenters can be ~1.4-2X more energy efficient than typical datacenters, and the ML-oriented accelerators inside them can be ~2-5X more effective than off-the-shelf systems. Remarkably, the choice of DNN, datacenter, and processor can reduce the carbon footprint up to ~100-1000X. These large factors also make retroactive estimates of energy cost difficult. To avoid miscalculations, we believe ML papers requiring large computational resources should make energy consumption and CO2e explicit when practical. We are working to be more transparent about energy use and CO2e in our future research. To help reduce the carbon footprint of ML, we believe energy usage and CO2e should be a key metric in evaluating models, and we are collaborating with MLPerf developers to include energy usage during training and inference in this industry standard benchmark."}48{"work_id": 4384697207, "title": "Methods and Techniques for Inter-cloud Data Transfers", "abstract": "This thesis focuses on methods and techniques for transferring data from a compute cloud to a storage cloud by using a buffer at the compute cloud.A space driven, a time driven, and a hybrid buffering technique are discussed.It was found that using buffering at the compute cloud is only beneficial when the load on the system is high.To improve performance a threshold-based and a machine learning based adaptive method are introduced at the buffer to dynamically change between buffering techniques at runtime based on the load on the system state.Using the adaptive methods leads to a significant improvement in performance.A proof-of-concept prototype was built for the different data transfer methods and techniques researched.Experiments performed on the prototype provide insights into system behaviour and performance."}49{"work_id": 2783525259, "title": "Loihi: A Neuromorphic Manycore Processor with On-Chip Learning", "abstract": "Loihi is a 60-mm2 chip fabricated in Intels 14-nm process that advances the state-of-the-art modeling of spiking neural networks in silicon. It integrates a wide range of novel features for the field, such as hierarchical connectivity, dendritic compartments, synaptic delays, and, most importantly, programmable synaptic learning rules. Running a spiking convolutional form of the Locally Competitive Algorithm, Loihi can solve LASSO optimization problems with over three orders of magnitude superior energy-delay-product compared to conventional solvers running on a CPU iso-process/voltage/area. This provides an unambiguous example of spike-based computation, outperforming all known conventional solutions."}50{"work_id": 7115707641, "title": "The carbon and water footprints of data centers and what this could mean for artificial intelligence", "abstract": "Although there are ways to estimate the global power demand of artificial intelligence (AI) systems, it remains challenging to quantify the associated carbon and water footprints. The lack of distinction between AI and non-AI workloads in the environmental reports of data center operators makes it possible to assess the environmental impact of AI workloads only by approximating them through data centers' general performance metrics. The environmental disclosure of tech companies is, however, often insufficient to determine even the total data center performance of these companies. The shortcomings in the environmental disclosure of data center operators could be remedied with new policies mandating the disclosure of additional metrics. Because the environmental impact of data centers is growing rapidly, the urgency of transparency in the tech sector is also increasing. The carbon footprint of AI systems alone could be between 32.6 and 79.7 million tons of CO 2 emissions in 2025, while the water footprint could reach 312.5\u2013764.6 billion L."}51{"work_id": 7130516837, "title": "Gaussian-Function-Driven Chaotic Memristive Dynamics with Application to Image Encryption", "abstract": "This paper develops a Gaussian-function-driven chaotic memristive system in which the magnetic flux is characterized by a Gaussian function, distinguishing it from prevailing formulations that employ piecewise, polynomial, or weakly smooth nonlinearities. The proposed system exhibits the key fingerprints of memristive behavior, including pinched hysteresis loops, local activity and nonvolatility. Integrating the Gaussian function into Chua\u2019s circuit produces a multiscroll chaotic Gaussian memristive system, where the attractors emerge in an organized and predictable pattern, with their numbers and spatial distribution determined by the Gaussian kernel parameters. To demonstrate the applicability of the proposed chaotic system, an image encryption scheme exploits the chaotic sequences generated by this system and integrates adaptive Zigzag scrambling, DNA encoding and bit-plane cross-diffusion. Experimental results show that the scheme achieves uniform histograms, negligible pixel correlation, strong key sensitivity and high resistance to noise and differential attacks. The incorporation of Gaussian function enriches the framework of chaotic memristive systems and establishes a flexible and robust foundation for secure image encryption."}52{"work_id": 3046751266, "title": "Scalable molecular dynamics on CPU and GPU architectures with NAMD", "abstract": "NAMDis a molecular dynamics program designed for high-performance simulations of very large biological objects on CPU- and GPU-based architectures. NAMD offers scalable performance on petascale parallel supercomputers consisting of hundreds of thousands of cores, as well as on inexpensive commodity clusters commonly found in academic environments. It is written in C++ and leans on Charm++ parallel objects for optimal performance on low-latency architectures. NAMD is a versatile, multipurpose code that gathers state-of-the-art algorithms to carry out simulations in apt thermodynamic ensembles, using the widely popular CHARMM, AMBER, OPLS, and GROMOS biomolecular force fields. Here, we review the main features of NAMD that allow both equilibrium and enhanced-sampling molecular dynamics simulations with numerical efficiency. We describe the underlying concepts utilized by NAMD and their implementation, most notably for handling long-range electrostatics; controlling the temperature, pressure, and pH; applying external potentials on tailored grids; leveraging massively parallel resources in multiple-copy simulations; and hybrid quantum-mechanical/molecular-mechanical descriptions. We detail the variety of options offered by NAMD for enhanced-sampling simulations aimed at determining free-energy differences of either alchemical or geometrical transformations and outline their applicability to specific problems. Last, we discuss the roadmap for the development of NAMD and our current efforts toward achieving optimal performance on GPU-based architectures, for pushing back the limitations that have prevented biologically realistic billion-atom objects to be fruitfully simulated, and for making large-scale simulations less expensive and easier to set up, run, and analyze. NAMD is distributed free of charge with its source code at www.ks.uiuc.edu."}53{"work_id": 4403458785, "title": "Artificial intelligence in higher education: exploring faculty use, self-efficacy, distinct profiles, and professional development needs", "abstract": "Abstract Faculty perspectives on the use of artificial intelligence (AI) in higher education are crucial for AI\u2019s meaningful integration into teaching and learning, yet research is scarce. This paper presents a study designed to gain insight into faculty members\u2019 ( N = 122) AI self-efficacy and distinct latent profiles, perceived benefits, challenges, use, and professional development needs related to AI. The respondents saw greater equity in education as AI\u2019s greatest benefit, while students and faculty members\u2019 lack of AI literacy was among the greatest challenges, with the majority interested in professional development. Latent class analysis revealed four distinct faculty member profiles: optimistic, critical, critically reflected, and neutral. The optimistic profile moderates the relationship between self-efficacy and usage. The development of adequate support services is suggested for successful and sustainable digital transformation."}54{"work_id": 4308245305, "title": "Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model", "abstract": "Progress in machine learning (ML) comes with a cost to the environment, given that training ML models requires significant computational resources, energy and materials. In the present article, we aim to quantify the carbon footprint of BLOOM, a 176-billion parameter language model, across its life cycle. We estimate that BLOOM's final training emitted approximately 24.7 tonnes of~\\carboneq~if we consider only the dynamic power consumption, and 50.5 tonnes if we account for all processes ranging from equipment manufacturing to energy-based operational consumption. We also study the energy requirements and carbon emissions of its deployment for inference via an API endpoint receiving user queries in real-time. We conclude with a discussion regarding the difficulty of precisely estimating the carbon footprint of ML models and future research directions that can contribute towards improving carbon emissions reporting."}55{"work_id": 1224547023, "title": "Model-Based Framework for Schedulability Analysis Using Uppaal 4.1", "abstract": "Embedded systems involve the monitoring and control of complex physical processes using applications running on dedicated execution platforms in a resource-constrained manner in terms of, for example, memory, processing power, bandwidth, energy consumption, and timing behavior."}56{"work_id": 4401304802, "title": "Best Practices Guide for Energy-Efficient Data Center Design", "abstract": "This guide provides an overview of best practices for energy-efficient data center design which spans the categories of information technology (IT) systems and their environmental conditions, data center air management, cooling and electrical systems, and heat recovery. IT system energy efficiency and environmental conditions are presented first because measures taken in these areas have a cascading effect of secondary energy savings for the mechanical and electrical systems. This guide concludes with a section on metrics and benchmarking values by which a data center and its systems energy efficiency can be evaluated. No design guide can offer \u201cthe most energy-efficient\u201d data center design but the guidelines that follow offer suggestions that provide efficiency benefits for a wide variety of data center scenarios."}57{"work_id": 4286840061, "title": "Algorithm-Architecture Co-Design for Domain-Specific Accelerators in Communication and Artificial Intelligence", "abstract": "The past decade has witnessed an explosive growth of data and the needs for high-speed data communications and processing. The needs continue to drive the development of new hardware for transmitting more data reliably and processing more data to obtain a higher level of intelligence. This thesis work explores algorithm-architecture co-design approach to derive efficient solutions for domain-specific communication and machine learning accelerators. It focuses on advanced and most compute-intensive accelerator designs for: 1) channel coding for data transmission, including polar codes and low-density parity-check (LDPC) codes, and 2) neural networks for machine learning, including differentiable neural computer (DNC) and neural ordinary differential equations (NODE). It also covers an interdisciplinary area of AI-aided communication, exploring DNC-aided flip decoding for polar codes. This work introduces a split-tree successive-cancellation list (SCL) decoder that works by dividing a polar code's decoding tree to sub-trees following a split-tree algorithm. Through algorithm-architecture co-optimizations, a 0.64mm2 40nm test chip implements a split-4, list-2, 8-frame-interleaved decoder that supports configurable code lengths up to 1024 bit and variable code rates. This work advances LDPC codes in two aspects: 1) we take the inspiration from simulated annealing to generalize the post-processor design using three methods: quenching, extended heating, and focused heating, each of which targets a different decoding error structure. The resulting post-processor is demonstrated to lower the error rate by two orders of magnitudes. 2) We also present a fully parallel decoder for a (160, 80) regular-(2, 4) NB-LDPC code over Galois field GF(64) in 65nm CMOS. The decoder employs fine-grained dynamic clock gating and decoder early termination to achieve a throughput of 1.22Gb/s and an energy efficiency of 3.03nJ/b. This work contributes to the hardware acceleration of DNC. We present HiMA, a tiled, history-based memory access engine with distributed memories in tiles. HiMA incorporates a traffic-aware multi-mode network-on-chip (NoC), an optimal submatrix-based memory partition and a two-stage usage sort method leveraging distributed tiles. We create a distributed DNC (DNC-D) to allow almost all memory operations to be applied to local memories. In a 40nm design, HiMA running DNC and DNC-D demonstrates 39.1x higher speed, 164.3x better area efficiency, and 61.2x better power efficiency over the state-of-the-art accelerators. This work contributes to the hardware acceleration of NODE for improved modeling capability of continuous-time events. We carry out algorithm-architecture co-design: 1) we propose adaptive neural activation sparsity for up to 80% complexity reduction while maintaining excellent training accuracy. 2) we develop a multi-mode PE design for NODE compute kernels with configurable interconnects between PEs to handle a variety of numerical ODE solvers. The hardware efficiency is further enhanced by exploring hardware reuse and hierarchical memory. Lastly, this work investigates an interdisciplinary area of AI-aided communication, applying DNC for flip decoding of polar codes. We develop new state and action encoding with a two-phase decoding flow. Simulation results show that proposed DNC-aided SCL-Flip (DNC-SCLF) decoding demonstrates up to 0.34dB coding gain improvement or 54.2% reduction in average number of decoding attempts compared to prior works. The five pieces of work presented in this thesis tackle hardware acceleration challenges in domain-specific computing using algorithm-architecture co-design techniques. The results of this work contribute to the developments of hardware accelerators in channel coding and deep learning, enabling next-generation communication and artificial intelligence from theory to efficient hardware."}58{"work_id": 7130433466, "title": "A Second-Order Memristor Method to Construct Memristive Neural Networks With Multi-Butterfly and Multi-Scroll Dynamics", "abstract": "Memristor-based Hopfield neural networks (MHNNs) exhibit rich chaotic dynamics and bear closer hardware resemblance to the biological brain, making them well suited for emulating neural dynamical behaviors. However, most existing MHNNs are constructed with first-order memristors. This paper proposes a novel second-order memristor (SOM) approach for constructing MHNNs with enriched chaotic dynamics. Specifically, a second-order memristor is incorporated into a three-neuron Hopfield neural network to emulate the magnetic coupling mechanism between neurons, thereby forming a second-order memristor-based neural network (SOM-HNN). Comprehensive dynamical analyses, including bifurcation diagrams, Lyapunov exponent spectra, and numerical simulations, confirm that the proposed SOM-HNN exhibits richer and more intricate chaos behaviors than its first-order counterparts. Remarkably, the proposed SOM-HNN can simultaneously generate butterfly and scroll attractors, multi-butterfly and multi-scroll attractors, as well as initial-boosed coexisting multi-butterfly and multi-scroll attractors, thereby substantially enhancing its dynamical diversity. To the best of our knowledge, this is the first report of both multi-butterfly and multi-scroll dynamics in a neural network. Furthermore, the SOM-HNN is implemented in hardware using analog circuits and a digital field-programmable gate array (FPGA) platform. Experimental results demonstrate the network\u2019s abundant dynamical features and its feasibility for efficient hardware realization in neuromorphic engineering applications."}59{"work_id": 7135102590, "title": "Biologically Plausible Memristive Decision-Making Circuit for Adaptive Control in Industrial Autonomous Navigation", "abstract": "In biological decision-making, adaptive behavior arises from the interaction between structured task context, expectation, action selection, and feedback-based learning. While most existing studies reproduce reward-driven responses under clear sensory stimuli, decision formation under weak or absent sensory evidence, such as low or 0% contrast conditions, remains insufficiently explored. To address this issue, this work proposes a biologically plausible memristive decision-making framework based on a block-structured task paradigm. The proposed system adopts a closed-loop architecture composed of four functional modules: stimulus, expectation, action, and reward/punishment. Sensory information is encoded when available, while the expectation pathway provides prior-guided modulation when sensory evidence becomes weak or unreliable. Action selection is generated through competitive integration, and reward\u2013punishment feedback dynamically corrects decision bias and reinforces appropriate responses. Through this hierarchical interaction, stable decision behavior can be achieved even in the absence of explicit sensory inputs. PSPICE simulations are conducted to analyze system dynamics and validate the corrective role of the reward\u2013punishment mechanism under weak and ambiguous conditions. In addition, the proposed framework is demonstrated in an industrial autonomous navigation scenario, illustrating its scalability and applicability for adaptive decision-making under uncertainty."}60{"work_id": 4394727812, "title": "Large-scale photonic chiplet Taichi empowers 160-TOPS/W artificial general intelligence", "abstract": "The pursuit of artificial general intelligence (AGI) continuously demands higher computing performance. Despite the superior processing speed and efficiency of integrated photonic circuits, their capacity and scalability are restricted by unavoidable errors, such that only simple tasks and shallow models are realized. To support modern AGIs, we designed Taichi-large-scale photonic chiplets based on an integrated diffractive-interference hybrid design and a general distributed computing architecture that has millions-of-neurons capability with 160-tera-operations per second per watt (TOPS/W) energy efficiency. Taichi experimentally achieved on-chip 1000-category-level classification (testing at 91.89% accuracy in the 1623-category Omniglot dataset) and high-fidelity artificial intelligence-generated content with up to two orders of magnitude of improvement in efficiency. Taichi paves the way for large-scale photonic computing and advanced tasks, further exploiting the flexibility and potential of photonics for modern AGI."}61{"work_id": 4385481791, "title": "Reducing the Carbon Impact of Generative AI Inference (today and in 2035)", "abstract": "Generative AI, exemplified in ChatGPT, Dall-E 2, and Stable Diffusion, are exciting new applications consuming growing quantities of computing. We study the compute, energy, and carbon impacts of generative AI inference. Using ChatGPT as an exemplar, we create a workload model and compare request direction approaches (Local, Balance, CarbonMin), assessing their power use and carbon impacts."}62{"work_id": 2921153466, "title": "Basic Performance Measurements of the Intel Optane DC Persistent Memory Module", "abstract": "Scalable nonvolatile memory DIMMs will finally be commercially available with the release of the Intel Optane DC Persistent Memory Module (or just \"Optane DC PMM\"). This new nonvolatile DIMM supports byte-granularity accesses with access times on the order of DRAM, while also providing data storage that survives power outages. This work comprises the first in-depth, scholarly, performance review of Intel's Optane DC PMM, exploring its capabilities as a main memory device, and as persistent, byte-addressable memory exposed to user-space applications. This report details the technologies performance under a number of modes and scenarios, and across a wide variety of macro-scale benchmarks. Optane DC PMMs can be used as large memory devices with a DRAM cache to hide their lower bandwidth and higher latency. When used in this Memory (or cached) mode, Optane DC memory has little impact on applications with small memory footprints. Applications with larger memory footprints may experience some slow-down relative to DRAM, but are now able to keep much more data in memory. When used under a file system, Optane DC PMMs can result in significant performance gains, especially when the file system is optimized to use the load/store interface of the Optane DC PMM and the application uses many small, persistent writes. For instance, using the NOVA-relaxed NVMM file system, we can improve the performance of Kyoto Cabinet by almost 2x. Optane DC PMMs can also enable user-space persistence where the application explicitly controls its writes into persistent Optane DC media. In our experiments, modified applications that used user-space Optane DC persistence generally outperformed their file system counterparts. For instance, the persistent version of RocksDB performed almost 2x faster than the equivalent program utilizing an NVMM-aware file system."}63{"work_id": 4283157527, "title": "Measuring the Carbon Intensity of AI in Cloud Instances", "abstract": "The advent of cloud computing has provided people around the world with unprecedented access to computational power and enabled rapid growth in technologies such as machine learning, the computational demands of which incur a high energy cost and a commensurate carbon footprint. As a result, recent scholarship has called for better estimates of the greenhouse gas impact of AI: data scientists today do not have easy or reliable access to measurements of this information, which precludes development of actionable tactics. We argue that cloud providers presenting information about software carbon intensity to users is a fundamental stepping stone towards minimizing emissions."}64{"work_id": 4396654316, "title": "Digital financial inclusion, financial stability, and sustainable development: Evidence from a quantile\u2010on\u2010quantile regression and wavelet coherence", "abstract": "Abstract The current research examines the influence of digital financial inclusion (DFI) and financial stability (FS) on sustainable development in Vietnam through a dataset spanning during 2004\u20132022. Non\u2010parametric methods, such as quantile regression, Granger tests in different quantiles, and wavelet analysis were employed to elucidate the effect of different quantiles of FS and DFI on various quantiles of sustainable development. Our primary contribution lies in integrating three research domains to examine the relationships among all three factors simultaneously, all within the framework of one of the biggest developing economies. Our findings exhibit the strongly positive influence of FS and DFI on sustainable development in Vietnam across various quantiles and frequencies. This suggests that FS and DFI support the transition of country toward sustainable development."}65{"work_id": 2781738013, "title": "Quantum Computing in the NISQ era and beyond", "abstract": "Noisy Intermediate-Scale Quantum (NISQ) technology will be available in the near future. Quantum computers with 50-100 qubits may be able to perform tasks which surpass the capabilities of today's classical digital computers, but noise in quantum gates will limit the size of quantum circuits that can be executed reliably. NISQ devices will be useful tools for exploring many-body quantum physics, and may have other useful applications, but the 100-qubit quantum computer will not change the world right away - we should regard it as a significant step toward the more powerful quantum technologies of the future. Quantum technologists should continue to strive for more accurate quantum gates and, eventually, fully fault-tolerant quantum computing."}66{"work_id": 3137672894, "title": "SSUR: An Approach to Optimizing Virtual Machine Allocation Strategy Based on User Requirements for Cloud Data Center", "abstract": "Cloud data centers provide services for an increasing number of applications. The virtual machines (VMs) that perform the corresponding application tasks need to be allocated to physical machines (PMs). For VM allocation, cloud service centers consider both energy consumption and quality of service (QoS), while cloud users are primarily concerned with their own needs, such as throughput and reliability. This paper proposes an allocation scheme for optimization based on user requirements in a cloud data center. First, various application requests from mobile phones and devices, which are regarded as a group of VM lists in the data center, are submitted to the cloud platform. Our method first allocates these arriving VMs to appropriate PMs based on their usage of hardware resources and the current throughput of the PMs in the data center. Second, due to dynamic workloads, the loads of the PMs that host these VMs may become very high. CPU utilization thresholds are set to determine whether migration is required, and the energy consumption before and after allocation is used to choose which VMs are reallocated. A suitable strategy for VM migration and PM shutdown can improve reliability and reduce energy consumption. Finally, it is shown through experimental simulations that compared with two existing algorithms, on the premise that the user requirements are met, the proposed method offers good performance in terms of total energy consumption, CPU utilization, number of PMs used and number of service-level agreement (SLA) violation."}67{"work_id": 2963840171, "title": "Information and Communications Technologies for Sustainable Development Goals: State-of-the-Art, Needs and Perspectives", "abstract": "In September 2015, the United Nations General Assembly accepted the 2030 Development Agenda, which has included 92 paragraphs, and the Paragraph 91 defined 17 sustainable development goals (SDGs) and 169 associated targets. The goal of this paper is to discover the correlations among SDGs and information and communications technologies (ICTs). This paper discusses the roles and opportunities that ICTs play in pursuing the SDGs. We identify a number of research gaps to those three pillars, social, economic, and environmental perspectives, of sustainable development. After extensive literature reviews on the SDG-related research initiatives and activities, we find that the majority of contributions to SDGs recognized by the IEEE and ACM research communities have mainly focused on the technical aspects while there are lack of the holistic social good perspectives. Therefore, there are essential and urgent needs to raise the awareness and call for attentions on how to innovate and energize ICTs in order to best assist all nations to achieve the SDGs by 2030."}68{"work_id": 4389373471, "title": "Logical quantum processor based on reconfigurable atom arrays", "abstract": "Abstract Suppressing errors is the central challenge for useful quantum computing 1 , requiring quantum error correction (QEC) 2\u20136 for large-scale processing. However, the overhead in the realization of error-corrected \u2018logical\u2019 qubits, in which information is encoded across many physical qubits for redundancy 2\u20134 , poses substantial challenges to large-scale logical quantum computing. Here we report the realization of a programmable quantum processor based on encoded logical qubits operating with up to 280 physical qubits. Using logical-level control and a zoned architecture in reconfigurable neutral-atom arrays 7 , our system combines high two-qubit gate fidelities 8 , arbitrary connectivity 7,9 , as well as fully programmable single-qubit rotations and mid-circuit readout 10\u201315 . Operating this logical processor with various types of encoding, we demonstrate improvement of a two-qubit logic gate by scaling surface-code 6 distance from d = 3 to d = 7, preparation of colour-code qubits with break-even fidelities 5 , fault-tolerant creation of logical Greenberger\u2013Horne\u2013Zeilinger (GHZ) states and feedforward entanglement teleportation, as well as operation of 40 colour-code qubits. Finally, using 3D [[8,3,2]] code blocks 16,17 , we realize computationally complex sampling circuits 18 with up to 48 logical qubits entangled with hypercube connectivity 19 with 228 logical two-qubit gates and 48 logical CCZ gates 20 . We find that this logical encoding substantially improves algorithmic performance with error detection, outperforming physical-qubit fidelities at both cross-entropy benchmarking and quantum simulations of fast scrambling 21,22 . These results herald the advent of early error-corrected quantum computation and chart a path towards large-scale logical processors."}69{"work_id": 2788945937, "title": "Classification with Quantum Neural Networks on Near Term Processors", "abstract": "We introduce a quantum neural network, QNN, that can represent labeled data, classical or quantum, and be trained by supervised learning. The quantum circuit consists of a sequence of parameter dependent unitary transformations which acts on an input quantum state. For binary classification a single Pauli operator is measured on a designated readout qubit. The measured output is the quantum neural network\u2019s predictor of the binary label of the input state. We show through classical simulation that parameters can be found that allow the QNN to learn to correctly distinguish the two data sets. We then discuss presenting the data as quantum superpositions of computational basis states corresponding to different label values. Here we show through simulation that learning is possible. We consider using our QNN to learn the label of a general quantum state. By example we show that this can be done. Our work is exploratory and relies on the classical simulation of small quantum systems. The QNN proposed here was designed with near-term quantum processors in mind. Therefore it will be possible to run this QNN on a near term gate model quantum computer where its power can be explored beyond what can be explored with simulation."}70{"work_id": 4386285839, "title": "Managing the Strategic Transformation of Higher Education through Artificial Intelligence", "abstract": "Considering the rapid advancements in artificial intelligence (AI) and their potential implications for the higher education sector, this article seeks to critically evaluate the strategic adoption of AI in the framework of \u201csmart universities\u201d. We envisage these innovative institutions as the imminent evolution in higher education, harnessing AI and quantum technologies to reshape academic and administrative processes. The core presumption is that through such integration, universities can achieve personalized learning trajectories, enhanced accessibility, economic efficiency, and a boost in overall operational performance. However, venturing into this new educational paradigm necessitates a thorough exploration of potential pitfalls, including questions surrounding educational quality, potential job losses, risks of bias, privacy breaches, and safety concerns. Our primary objective is to offer a balanced assessment to aid stakeholders in making informed strategic decisions about endorsing and advancing the smart university model. A pivotal factor in this discourse is the acceptance of qualifications from AI-enriched institutions by employers, a variable that may drastically redefine the education sector\u2019s trajectory. Within the context of a comprehensive analysis of its broader societal impact, this article also delves into the ramifications of AI-driven innovations for historically Black colleges and universities (HBCUs)."}71{"work_id": 3099878876, "title": "Array programming with NumPy", "abstract": "Abstract Array programming provides a powerful, compact and expressive syntax for accessing, manipulating and operating on data in vectors, matrices and higher-dimensional arrays. NumPy is the primary array programming library for the Python language. It has an essential role in research analysis pipelines in fields as diverse as physics, chemistry, astronomy, geoscience, biology, psychology, materials science, engineering, finance and economics. For example, in astronomy, NumPy was an important part of the software stack used in the discovery of gravitational waves1 and in the first imaging of a black hole2. Here we review how a few fundamental array concepts lead to a simple and powerful programming paradigm for organizing, exploring and analysing scientific data. NumPy is the foundation upon which the scientific Python ecosystem is constructed. It is so pervasive that several projects, targeting audiences with specialized needs, have developed their own NumPy-like interfaces and array objects. Owing to its central position in the ecosystem, NumPy increasingly acts as an interoperability layer between such array computation libraries and, together with its application programming interface (API), provides a flexible framework to support the next decade of scientific and industrial analysis."}72{"work_id": 4366587887, "title": "Promising Emerging Technologies for Teaching and Learning: Recent Developments and Future Challenges", "abstract": "As time goes on and the number of people who use information and communication technology (ICT) grows, emerging technologies are receiving a lot of attention from academics, researchers, and users. Emerging technologies are changing more quickly than ever, and it is important to start building skills early in education by using the right examples to teach and learn about digital citizenship. New pedagogies support individual teaching and learning methodologies for knowledge acquisition and self-directed learning. Hundreds of digital education tools have been developed to increase student autonomy, enhance academic process management, promote collaboration, and improve communication between teachers and students. This article examines the current state of emerging technologies from a variety of educational viewpoints, highlights a few of them, and discusses both their established and potential educational benefits. Additionally, it offers in-depth debates on recent developments and future challenges from a wide range of perspectives. The analysis focuses on peer-reviewed research articles published in well-renowned publications databases over the last 8 years, drawing upon a bibliometric analysis using VOSviewer. The results of this study are also anticipated to serve as a substantial impetus for other universities and colleges worldwide to utilize innovatively and revolutionized technology for better academic performance."}73{"work_id": 4409126690, "title": "IBM quantum computers: evolution, performance, and future directions", "abstract": "Abstract Quantum computers represent a transformative frontier in computational technology, promising exponential speedups beyond classical computing limits. IBM Quantum has led significant advancements in both hardware and software, providing access to quantum hardware via IBM Cloud\u00ae since 2016 and achieving a milestone with the world\u2019s first accessible quantum computer. This paper explores IBM\u2019s journey in quantum computing, focusing on its contributions to both hardware and software, as well as the development of practical quantum computers. We trace the evolution of IBM Quantum\u2019s processors, from the early canary processors to the milestone of surpassing the 1000-qubit barrier. In addition to these technological strides, we delve into the practical applications of quantum computing, particularly within nine key industries: airlines, banking, healthcare, electronics, life sciences, and more. We also explore IBM Quantum\u2019s case studies and strategic partnerships with organizations such as Boeing, CERN, ExxonMobil, and Cleveland Clinic, which are helping to bridge the gap between theoretical research and real-world applications. Further, we examine the key challenges and solutions in scaling quantum systems and achieving fault tolerance, highlighting IBM\u2019s efforts toward building practical, fault-tolerant quantum systems capable of addressing real-world problems."}74{"work_id": 4406693713, "title": "Scaling and networking a modular photonic quantum computer", "abstract": "Photonics offers a promising platform for quantum computing1\u20134, owing to the availability of chip integration for mass-manufacturable modules, fibre optics for networking and room-temperature operation of most components. However, experimental demonstrations are needed of complete integrated systems comprising all basic functionalities for universal and fault-tolerant operation5. Here we construct a (sub-performant) scale model of a quantum computer using 35 photonic chips to demonstrate its functionality and feasibility. This combines all the primitive components as discrete, scalable rack-deployed modules networked over fibre-optic interconnects, including 84 squeezers6 and 36 photon-number-resolving detectors furnishing 12 physical qubit modes at each clock cycle. We use this machine, which we name Aurora, to synthesize a cluster state7 entangled across separate chips with 86.4 billion modes, and demonstrate its capability of implementing the foliated distance-2 repetition code with real-time decoding. The key building blocks needed for universality and fault tolerance are demonstrated: heralded synthesis of single-temporal-mode non-Gaussian resource states, real-time multiplexing actuated on photon-number-resolving detection, spatiotemporal cluster-state formation with fibre buffers, and adaptive measurements implemented using chip-integrated homodyne detectors with real-time single-clock-cycle feedforward. We also present a detailed analysis of our architecture\u2019s tolerances for optical loss, which is the dominant and most challenging hurdle to crossing the fault-tolerant threshold. This work lays out the path to cross the fault-tolerant threshold and scale photonic quantum computers to the point of addressing useful applications. A proof-of-principle study reports a complete photonic quantum computer architecture that can, once appropriate component performance is achieved, deliver a universal and fault-tolerant quantum computer."}75{"work_id": 4380627387, "title": "Evidence for the utility of quantum computing before fault tolerance", "abstract": "Abstract Quantum computing promises to offer substantial speed-ups over its classical counterpart for certain problems. However, the greatest impediment to realizing its full potential is noise that is inherent to these systems. The widely accepted solution to this challenge is the implementation of fault-tolerant quantum circuits, which is out of reach for current processors. Here we report experiments on a noisy 127-qubit processor and demonstrate the measurement of accurate expectation values for circuit volumes at a scale beyond brute-force classical computation. We argue that this represents evidence for the utility of quantum computing in a pre-fault-tolerant era. These experimental results are enabled by advances in the coherence and calibration of a superconducting processor at this scale and the ability to characterize 1 and controllably manipulate noise across such a large device. We establish the accuracy of the measured expectation values by comparing them with the output of exactly verifiable circuits. In the regime of strong entanglement, the quantum computer provides correct results for which leading classical approximations such as pure-state-based 1D (matrix product states, MPS) and 2D (isometric tensor network states, isoTNS) tensor network methods 2,3 break down. These experiments demonstrate a foundational tool for the realization of near-term quantum applications 4,5 ."}76{"work_id": 7140797778, "title": "Carbon Footprint Efficiency Ratio: A Unified Indicator for Green Evaluation of Evolutionary Algorithms", "abstract": "Evolutionary algorithms (EAs) are widely adopted for black-box optimization, yet evaluation protocols remain largely accuracy-centric and often omit computational energy use, hindering carbon-aware deployment in green computing settings. We propose the Carbon Footprint Efficiency Ratio (CFER) to quantify optimization benefit per unit carbon emission and enable a unified assessment of carbon efficiency. To compute CFER without dedicated power-monitoring hardware, we develop a device-independent energy estimation model based on the Multimodal Energy Consumption Factor (MECF). Using a dataset constructed from 13 benchmark functions and 8 classical EAs, MECF is identified via outlier removal and Huber regression, achieving a mean estimation error of 1.80% and a maximum error of 2.51%. On the CEC2020 test suite, CFER-based comparisons reveal rankings that differ substantially from accuracy-only evaluations, highlighting algorithms that improve carbon efficiency with only marginal accuracy loss. Beyond metric-based assessment, we introduce a CFER-aware green optimization mechanism as an algorithm-agnostic plug-in control layer. Driven by real-time carbon estimation and marginal CFER feedback, it adaptively budgets computation through carbon-budget termination and population shrinking under low marginal carbon efficiency. Integrated into GA, DE, and PSO for uncrewed aerial vehicle (UAV) three-dimensional path planning, the resulting CFER-aware variants reduce CO2emissions by 45.91%, 28.28%, and 59.14%, respectively, while increasing CFER by 78.59%, 32.45%, and 133.70%, respectively, yielding an improved performance\u2013carbon trade-off with minor performance-score degradation. Overall, the proposed framework supports carbon-aware evaluation, algorithm selection, and adaptive computation budgeting for green AI systems."}77{"work_id": 7130515337, "title": "Artificial intelligence literacy and academic resilience in undergraduate nursing students: the mediating role of self-efficacy and artificial intelligence anxiety", "abstract": "BACKGROUND: Artificial intelligence is increasingly integrated into nursing education, making AI literacy a key competency for students. Academic resilience is also essential for coping with learning challenges. However, the mechanisms linking AI literacy to academic resilience remain unclear, particularly the roles of self-efficacy and AI anxiety. This study examines these associations and tests the mediating effects of self-efficacy and AI anxiety. METHODS: A cross-sectional study was conducted among 415 undergraduate nursing students from a university in northern China. AI literacy, academic resilience, self-efficacy and AI anxiety were assessed using validated scales. Statistical analyses included descriptive analysis, Pearson correlation and serial mediation testing using the PROCESS macro in SPSS with bootstrap sampling. RESULTS: (1) Academic resilience was significantly and positively correlated with AI literacy (r = 0.229, p < 0.01) and self-efficacy (r = 0.340, p < 0.01), and negatively correlated with AI anxiety (r = -0.426, p < 0.01). AI literacy was positively correlated with self-efficacy (r = 0.118, p < 0.05) and negatively correlated with AI anxiety (r = -0.143, p < 0.01). (2) AI literacy was positively associated with self-efficacy (\u03b2 = 0.118, p = 0.017) and was negatively associated with AI anxiety (\u03b2 = -0.143, p = 0.006). Self-efficacy positively was positively associated with academic resilience (\u03b2 = 0.340, p < 0.001), while AI anxiety was negatively associated with academic resilience (\u03b2 = -0.426, p < 0.001). (3) In the model of AI literacy \u2192 self-efficacy \u2192 AI anxiety \u2192 academic resilience, the total indirect effect was 0.055. The mediating effect of self-efficacy accounted for 36.4% of the total indirect effect (0.020), while AI anxiety accounted for 58.2% (0.032). The serial mediation effect of self-efficacy and AI anxiety contributed 5.4% (0.003) to the total indirect effect, which was statistically significant but small in magnitude. Overall, the mediation pattern supported partial (rather than full) mediation. CONCLUSION: AI literacy is is an important correlate of nursing students\u2019 academic resilience. Strengthening AI literacy, improving self-efficacy and reducing AI anxiety may be linked to better adaptation to AI-supported learning environments. CLINICAL TRIAL NUMBER: Not applicable."}78{"work_id": 4407614581, "title": "Examining Teaching Competencies and Challenges While Integrating Artificial Intelligence in Higher Education", "abstract": "Abstract The rapid development of artificial intelligence (AI) technologies has demonstrated their affordances and limitations in revolutionizing pedagogical strategies in higher education. Given the lack of guidelines, policies, and resources to assist instructors in efficiently and ethically integrating AI into teaching and learning practices, this systematic review aimed to investigate AI integration competencies and challenges in higher education from the intelligent Technological Pedagogical Content Knowledge (TPACK) perspective. We first applied the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) to identify 23 studies published between 2019 and 2023 that met the inclusion and exclusion criteria. After conducting open coding and thematic analysis, the research findings showed four AI integration competencies and strategies, including 1) AI literacy and readiness as intelligent technological knowledge, 2) AI-supported innovative pedagogy to supplement instructor social presence and transform the instructor-student relationship, 3) AI as intelligent learning partners to increase student engagement in self-regulated learning and higher-order thinking skills, and 4) AI-driven learning experience design and delivery. AI integration limitations and challenges contained: 1) the short-term novelty effect, 2) digital divides, 3) technical deficiencies, and 4) ethical concerns. Based on the implications for future practices, a diagram was developed to illustrate the systematic considerations to support the sustainability of AI-assisted teaching and learning in higher education."}79{"work_id": 4394998532, "title": "PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation", "abstract": "This paper introduces two extensions to the popular PyTorch machine learning framework, TorchDynamo and TorchInductor, which implement the torch.compile feature released in PyTorch 2. TorchDynamo is a Python-level just-in-time (JIT) compiler that enables graph compilation in PyTorch programs without sacrificing the flexibility of Python. It achieves this by dynamically modifying Python bytecode before execution and extracting sequences of PyTorch operations into an FX graph, which is then JIT compiled using one of many extensible backends. TorchInductor is the default compiler backend for TorchDynamo, which translates PyTorch programs into OpenAI's Triton for GPUs and C++ for CPUs. Results show that TorchDynamo is able to capture graphs more robustly than prior approaches while adding minimal overhead, and TorchInductor is able to provide a 2.27\u00d7 inference and 1.41\u00d7 training geometric mean speedup on an NVIDIA A100 GPU across 180+ real-world models, which outperforms six other compilers. These extensions provide a new way to apply optimizations through compilers in eager mode frameworks like PyTorch."}80{"work_id": 2807211499, "title": "Towards Efficient and Flexible Object Storage Using Resource and Functional Partitioning", "abstract": "Modern storage systems are designed to manage data without considering the dynamicity of user or resource requirements. This design approach does not consider the complexities of the dynamically changing runtime application behaviors as well as the unique features of underlying resources. To this end, this dissertation studies how resource and functional partitioning strategies can improve efficiency and flexibility of object stores. This dissertation presents a series of practical and efficient techniques, algorithms, and optimizations to realize efficient and flexible object stores. The experimental evaluation demonstrates the effectiveness of our design choices and strategies to make object stores flexible and resource-aware."}81{"work_id": 2952332632, "title": "Towards a Reliable and Context-Based System Architecture for Autonomous Vehicles", "abstract": "Full vehicle autonomy excludes a takeover by passengers in case a safety-critical application fails. Therefore, the system responsible for operating the autonomous vehicle has to detect and handle failures autonomously. Moreover, this system has to ensure the safety of the passengers, as well as the safety of other road users at any given time. Especially in the initial phase of autonomous vehicles, building up consumer confidence is essential. Therefore, in this regard, handling all failures by simply performing an emergency stop is not desirable. In this paper, we introduce an approach enabling a dynamic and safe reconfiguration of the autonomous driving system to handle occurring hardware and software failures. Since the requirements concerning safe reconfiguration actions are significantly affected by the current context the car is experiencing, the developed reconfiguration approach is sensitive to context changes. Our approach defines three interconnected layers, which are distinguished by their level of awareness. The top layer, referred to as the context layer, is responsible for observing the context. These context observations, in turn, imply a set of requirements, which constitute the input for the reconfiguration layer. The latter layer is required to determine reconfiguration actions, which are then executed by the architecture layer."}82{"work_id": 2798337732, "title": "Reducing DRAM footprint with NVM in Facebook", "abstract": "Popular SSD-based key-value stores consume a large amount of DRAM in order to provide high-performance database operations. However, DRAM can be expensive for data center providers, especially given recent global supply shortages that have resulted in increasing DRAM costs. In this work, we design a key-value store, MyNVM, which leverages an NVM block device to reduce DRAM usage, and to reduce the total cost of ownership, while providing comparable latency and queries-per-second (QPS) as MyRocks on a server with a much larger amount of DRAM. Replacing DRAM with NVM introduces several challenges. In particular, NVM has limited read bandwidth, and it wears out quickly under a high write bandwidth."}83{"work_id": 3121141908, "title": "An optical neural chip for implementing complex-valued neural network", "abstract": "Complex-valued neural networks have many advantages over their real-valued counterparts. Conventional digital electronic computing platforms are incapable of executing truly complex-valued representations and operations. In contrast, optical computing platforms that encode information in both phase and magnitude can execute complex arithmetic by optical interference, offering significantly enhanced computational speed and energy efficiency. However, to date, most demonstrations of optical neural networks still only utilize conventional real-valued frameworks that are designed for digital computers, forfeiting many of the advantages of optical computing such as efficient complex-valued operations. In this article, we highlight an optical neural chip (ONC) that implements truly complex-valued neural networks. We benchmark the performance of our complex-valued ONC in four settings: simple Boolean tasks, species classification of an Iris dataset, classifying nonlinear datasets (Circle and Spiral), and handwriting recognition. Strong learning capabilities (i.e., high accuracy, fast convergence and the capability to construct nonlinear decision boundaries) are achieved by our complex-valued ONC compared to its real-valued counterpart."}84{"work_id": 4405183524, "title": "Quantum error correction below the surface code threshold", "abstract": "Quantum error correction [1, 2, 3, 4] provides a path to reach practical quantum computing by combining multiple physical qubits into a logical qubit, where the logical error rate is suppressed exponentially as more qubits are added. However, this exponential suppression only occurs if the physical error rate is below a critical threshold. Here, we present two below-threshold surface code memories on our newest generation of superconducting processors, Willow: a distance-7 code, and a distance-5 code integrated with a real-time decoder. The logical error rate of our larger quantum memory is suppressed by a factor of $$\\varLambda \\mathrm{=2.14}\\pm 0.02$$ when increasing the code distance by two, culminating in a 101-qubit distance-7 code with 0.143% \u00b1 0.003% error per cycle of error correction. This logical memory is also beyond break-even, exceeding its best physical qubit\u2019s lifetime by a factor of $$2.4\\pm 0.3$$ . Our system maintains below-threshold performance when decoding in real time, achieving an average decoder latency of 63 \u03bcs at distance-5 up to a million cycles, with a cycle time of 1.1 \u03bcs. We also run repetition codes up to distance-29 and find that logical performance is limited by rare correlated error events occurring approximately once every hour, or $$3$$ $$\\times $$ $${10}^{9}$$ cycles. Our results present device performance that, if scaled, could realize the operational requirements of large scale fault-tolerant quantum algorithms."}85{"work_id": 3182301081, "title": "Zico: Efficient {GPU} Memory Sharing for Concurrent {DNN} Training", "abstract": "GPUs are the workhorse in modern server infrastructure fueling advances in a number of compute-intensive workloads such as deep neural network (DNN) training. Several recent works propose solutions on sharing GPU resources across multiple concurrent DNN training jobs, but none of them address rapidly increasing memory footprint introduced by such job co-locations, which greatly limit the effectiveness of sharing GPU resources. In this paper, we present Zico, the first DNN system that aims at reducing the system-wide memory consumption for concurrent training. Zico keeps track of the memory usage pattern of individual training job by monitoring its progress on GPU computations and makes memory reclaimed from the job globally sharable. Based on this memory management scheme, Zico automatically decides a strategy to share memory among concurrent jobs with minimum delay on training while not exceeding a given memory budget such as GPU memory capacity. Our evaluation shows that Zico outperforms existing GPU sharing approaches and delivers benefits over a variety of job co-location scenarios."}86{"work_id": 7128511721, "title": "A Memristor-Based Neural Network Circuit With Classical Conditioning and Fear Generalization", "abstract": "The new generation of artificial intelligence (AI) is evolving from perception-oriented processing toward cognition-oriented intelligence, emphasizing learning, memory, decision-making, and adaptive behavior inspired by the human brain. In this paper, a memristor-based perceptual decision-making neural network circuit (PDNNC) is proposed to integrate multiple classical conditioning mechanisms and fear generalization within a unified hardware framework. Unlike existing studies that focus on isolated learning behaviors, the proposed circuit architecture simultaneously realizes associative learning, forgetting, latent inhibition, blocking effects, secondary conditioning, and emotion-modulated decision-making at the circuit level. A voltage-controlled threshold-type memristor model is employed to emulate synaptic plasticity, while modular perceptual and decision-making subcircuits are designed to ensure scalability and functional consistency. The complete circuit implementation is verified using PSPICE circuit simulations. Simulation results demonstrate the correct realization of diverse classical conditioning behaviors, controllable fear generalization, and adaptive decision responses under different perceptual stimuli. The proposed bio-inspired circuit has vast potential applications in realizing cognitive intelligence and robotics."}87{"work_id": 4297813615, "title": "DeepTrust^RT: Confidential Deep Neural Inference Meets Real-Time!", "abstract": "Deep Neural Networks (DNNs) are becoming common in \"learning-enabled\" time-critical applications such as autonomous driving and robotics. One approach to protect DNN inference from adversarial actions and preserve model privacy/confidentiality is to execute them within trusted enclaves available in modern processors. However, running DNN inference inside limited-capacity enclaves while ensuring timing guarantees is challenging due to (a) large size of DNN workloads and (b) extra switching between \"normal\" and \"trusted\" execution modes. This paper introduces new time-aware scheduling schemes - DeepTrust^RT - to securely execute deep neural inferences for learning-enabled real-time systems. We first propose a variant of EDF (called DeepTrust^RT-LW) that slices each DNN layer and runs them sequentially in the enclave. However, due to extra context switch overheads of individual layer slices, we further introduce a novel layer fusion technique (named DeepTrust^RT-FUSION). Our proposed scheme provides hard real-time guarantees by fusing multiple layers of DNN workload from multiple tasks; thus allowing them to fit and run concurrently within the enclaves while maintaining real-time guarantees. We implemented and tested DeepTrust^RT ideas on the Raspberry Pi platform running OP-TEE+DarkNet-TZ DNN APIs and three DNN workloads (AlexNet-squeezed, Tiny Darknet, YOLOv3-tiny). Compared to the layer-wise partitioning approach (DeepTrust^RT-LW), DeepTrust^RT-FUSION can schedule up to 3x more tasksets and reduce context switches by up to 11.12x. We further demonstrate the efficacy of DeepTrust^RT using a flight controller (ArduPilot) case study and find that DeepTrust^RT-FUSION retains real-time guarantees where DeepTrust^RT-LW becomes unschedulable."}88{"work_id": 4396986598, "title": "Quantum computing with Qiskit", "abstract": "We describe Qiskit, a software development kit for quantum information science. We discuss the key design decisions that have shaped its development, and examine the software architecture and its core components. We demonstrate an end-to-end workflow for solving a problem in condensed matter physics on a quantum computer that serves to highlight some of Qiskit's capabilities, for example the representation and optimization of circuits at various abstraction levels, its scalability and retargetability to new gates, and the use of quantum-classical computations via dynamic circuits. Lastly, we discuss some of the ecosystem of tools and plugins that extend Qiskit for various tasks, and the future ahead."}89{"work_id": 2156094048, "title": "Reverend Bayes on Inference Engines: A Distributed Hierarchical Approach", "abstract": "This paper presents generalizations of Bayes likelihood-ratio updating rule which facilitate an asynchronous propagation of the impacts of new beliefs and/or new evidence in hierarchically organized inference structures with multi-hypotheses variables. The computational scheme proposed specifies a set of belief parameters, communication messages and updating rules which guarantee that the diffusion of updated beliefs is accomplished in a single pass and complies with the tenets of Bayes calculus."}90{"work_id": 4295093682, "title": "fast16 as a Second-Order Subversion Attack", "abstract": "SentinelOne's April 2026 disclosure of fast16, a precision sabotage framework compiled in August 2005, predates Stuxnet by five years and corrupts floating-point calculations in engineering simulation software via kernel-mode filesystem filtering. The infosec community described the mechanism accurately but did not classify the attack class. This paper applies the nth-order attackframework from Bilar (NATO CCDCOE, 2009) and argues fast16 is a second-order subversion attack: the high-precision research end system is reached through two successive ancillary systems, the OS kernel filesystem stack and the engineering simulation software above it. The wormlet is a delivery mechanism, not a distinct ancillary hop. fast16 is structurally comparable to Thompson's compiler trojan but one order lower and more covert: it operates at runtime below the application layer with no source-level or on-disk trace, defeating the reproducible-build defense that catches Thompson. Stuxnet, dr0wned, and FLAW3D are mapped at the same second order in different substrates, showing the taxonomy holds across software and physical manufacturing pipelines. The classification yields concrete defenses (hardware-rooted driver attestation, diverse independent toolchains across validation hosts, air-gapped reference baselines) that follow from the framework rather than the incident report.Companion to Bilar (2026), \"Volt Typhoon as Operational Realization of the nth-Order Bleeding Thesis\" (Zenodo 19739954)."}91{"work_id": 4410153295, "title": "Bursting Firings in Memristive Hopfield Neural Network With Image Encryption and Hardware Implementation", "abstract": "By integrating memristors into a Hopfield neural network (HNN), a diverse range of dynamical behavior can be generated, which has significant implications for modeling and biomimetic applications of artificial neurons. However, research on the firing dynamics of HNNs remains relatively limited. In response, a memristive tri-neurons Hopfield neural network (MTN-HNN) was constructed, with the synapse of the second neuron replaced by the proposed memristor. A theoretical and experimental investigation of the dynamics of this neural network was conducted using general analytical tools, such as phase diagrams, Lyapunov exponents, bifurcation diagrams, and others. Experimental results indicate that the dynamics of the MTN-HNN is influenced by the internal parameters of the memristor, enabling the network to extend attractors in up to two directions and thereby form grid multi-scrolls. Notably, the MTN-HNN exhibits various firing modes, including periodic and chaotic bursting. Finally, an encryption scheme was proposed to demonstrate the potential of the MTN-HNN, and both the custom digital circuits and the encryption scheme were successfully implemented on a Field-Programmable Gate Array (FPGA)."}92{"work_id": 6922095860, "title": "Practical Mechanisms for Reducing Processor\u2013Memory Data Movement in Modern Workloads", "abstract": "Data movement between the memory system and computation units is one of the most critical challenges in designing high performance and energy-efficient computing systems. The high cost of data movement is forcing architects to rethink the fundamental design of computer systems. Recent advances in memory design enable the opportunity for architects to avoid unnecessary datamovement by performing processing-in-memory (PIM), also known as near-data processing (NDP). While PIM can allow many data-intensive applications to avoid moving data from memory to the CPU, it introduces new challenges for system architects and programmers. Our goal in this thesisis to make PIM effective and practical in conventional computing systems. Toward this end, this thesis presents three major directions: (1) examining the suitability of PIM across key workloads, (2) addressing major system challenges for adopting PIM in computing systems, and (3) redesigning applications aware of PIM capability. In line with these three major directions, we propose a seriesof practical mechanisms to reduce processor\u2013memory data movement in modern workloads: First, we comprehensively analyze the energy and performance impact of data movement for several widely-used Google consumer workloads. We find that PIM can significantly reduce datamovement for all of these workloads, by performing part of the computation close to memory. Each workload contains simple primitives and functions that contribute to a significant amount of the overall data movement. We investigate whether these primitives and functions are feasible to implement using PIM, given the limited area and power constraints of consumer devices. Our analysis shows that offloading these primitives to PIM logic, consisting of either simple cores or specialized accelerators, eliminates a large amount of data movement, and significantly reducestotal system energy execution time. Second, we address one of the key system challenges for communication with PIM logic by proposing an efficient cache coherence support for near-data accelerators (NDAs). We find that enforcing coherence with the rest of the system, which is already a major challenge for on-chip accelerators, becomes more difficult for NDAs. This is because (1) the cost of communication between NDAs and CPUs is high, and (2) NDA applications generate a large amount of off-chip datamovement. As a result, as we show in this work, existing coherence mechanisms eliminate most of the benefits of NDAs. Based on our observations, we propose CoNDA, a coherence mechanism that lets an NDA optimistically execute an NDA kernel, under the assumption that the NDA has all necessary coherence permissions. This optimistic execution allows CoNDA to gather information on the memory accesses performed by the NDA and by the rest of the system. CoNDA exploits this information to avoid performing unnecessary coherence requests, and thus, significantly reducesdata movement for coherence. We show that CoNDA significantly improves performance andreduces energy consumption compared to prior coherence mechanisms. Third, we propose a hardware\u2013software co-design approach aware of PIM for edge machinelearning (ML) accelerators to enable energy-efficient and high-performance inference execution. We analyze a commercial Edge TPU (tensor processing unit) using 24 Google edge neural network (NN) models (including CNNs, LSTMs, transducers, and RCNNs), and find that the accelerator suffers from three shortcomings, in terms of computational throughput, energy efficiency, andmemory access handling. We comprehensively study the characteristics of each NN layer in all of the Google edge models, and find that these shortcomings arise from the one-size-fits-all approach of the accelerator, as there is a high amount of heterogeneity in key layer characteristics both across different models and across different layers in the same model. To combat this inefficiency, we propose a new acceleration framework called Mensa. Mensa incorporates multiple heterogeneous ML edge accelerators (including both on-chip and near-data accelerators), each of which caters to the characteristics of a particular subset of models. At runtime, Mensa schedules each layer torun on the best-suited accelerator, accounting for both efficiency and inter-layer dependencies. We show that Mensa significantly improves inference energy and throughput, while reducing hardware cost and improving area efficiency over the Edge TPU and Eyeriss v2, two state-of-the-art edge ML accelerators. Lastly, we propose to redesign emerging modern hybrid databases to be aware of PIM capability, to enable real-time analysis. Hybrid transactional and analytical processing (HTAP) databasesystems can support real-time data analysis without the high costs of synchronizing across separate single-purpose databases. Unfortunately, for many applications that perform a high rate of data updates, state-of-the-art HTAP systems incur significant drops in transactional and/or analyticalthroughput compared to performing only transactions or only analytics in isolation, due to (1) data movement between the CPU and memory, (2) data update propagation, and (3) consistency costs. We propose Polynesia, a hardware\u2013software co-designed system for in-memory HTAP databases. Polynesia (1) divides the HTAP system into transactional and analytical processing islands, (2) implements custom algorithms and hardware to reduce the costs of update propagation and consistency, and (3) exploits processing-in-memory for the analytical islands to alleviate data movement. We show that Polynesia significantly outperforms three state-of-the-art HTAP systems and reduces energy consumption."}93{"work_id": 4415057924, "title": "Electricity demand and grid impacts of AI data centers: Challenges and prospects", "abstract": "The rapid growth of artificial intelligence (AI) is driving an unprecedented increase in the electricity demand of AI data centers, raising emerging challenges for electric power grids. Understanding the characteristics of AI data center loads and their interactions with the grid is therefore critical for ensuring both reliable power system operation and sustainable AI development. This paper provides a comprehensive review and vision of this evolving landscape. Specifically, this paper (i) presents an overview of AI data center infrastructure and its key components, (ii) examines the key characteristics and patterns of electricity demand across the stages of model preparation, training, fine-tuning, and inference, (iii) analyzes the critical challenges that AI data center loads pose to power systems across three interrelated timescales, including long-term planning and interconnection, short-term operation and electricity markets, and real-time dynamics and stability, and (iv) discusses potential solutions from the perspectives of the grid, AI data centers, and AI end-users to address these challenges. By synthesizing current knowledge and outlining future directions, this review aims to guide research and development in support of the joint advancement of AI data centers and power systems toward reliable, efficient, and sustainable operation."}94{"work_id": 7131231908, "title": "Reconfigurable Multiscroll Memristive Neural Network With Application to Telemedicine Privacy Protection", "abstract": "Constructing memristive neural networks (MNNs) with multiscroll chaotic attractors helps advance both theoretical and applied research on neural networks. However, the existing models mainly utilize complex memristor models with polynomial functions, nested composite functions, and so on, to generate multiscroll chaotic attractors, which leads to increased model complexity and difficulties in on-demand adjustment. Hence, this article proposes a reconfigurable multiscroll MNN (RMMNN) that can yield different types of multiscroll chaotic attractors merely by altering the memristive parameters without modifying its model. Through numerical methods, the complex dynamics of the RMMNN in different cases are analyzed, such as parameter-controlled multiscroll chaotic attractors, adjustable multistability, and parameter-induced transitions of multistability. In addition, the reliability of the numerical analysis is verified via the hardware circuits. Moreover, to address the issues of image security and low quality in telemedicine, a bidirectional rotation medical image encryption scheme (BRMIES) is developed based on the good pseudorandom chaotic sequences generated by RMMNN. Performance analysis demonstrates that BRMIES can effectively protect medical image and robustly handle various potential adverse interferences within telemedicine process."}95{"work_id": 4407217670, "title": "vAttention: Dynamic Memory Management for Serving LLMs without PagedAttention", "abstract": "PagedAttention is a popular approach for dynamic memory allocation in LLM serving systems. It enables on-demand allocation of GPU memory to mitigate KV cache fragmentation - a phenomenon that crippled the batch size (and consequently throughput) in prior systems. However, in trying to allocate physical memory at runtime, PagedAttention ends up changing the virtual memory layout of the KV cache from contiguous to non-contiguous. Such a design leads to non-trivial programming and performance overheads."}96{"work_id": 4390660405, "title": "The Optimization of Carbon Emission Prediction in Low Carbon Energy Economy Under Big Data", "abstract": "With the intensification of global climate change, low-carbon energy has become a hot topic, and governments around the world are implementing corresponding policies to promote its use. This research first establishes a Multi-universe Quantum Harmony Search-Algorithm Dynamic Fuzzy System Ensemble (MUQHS-DMFSE) composite model for carbon emission prediction. This model combines the MUQHS algorithm with the DMFSE method by designing the workflow of the MUQHS algorithm, creating a DMFSE composite prediction model, introducing a sliding factor matrix, and using the MUQHS algorithm to search for the optimal sliding factors, thus obtaining optimized prediction values. In the research on low-carbon economic development, the research applies the Data Envelopment Analysis (DEA) method and establishes Charnes-Cooper-Rhodes (CCR) and Banker-Charnes-Cooper (BCC) models to assess the technical efficiency, pure technical efficiency, and scale efficiency of decision-making units. This research also uses the BCC model to project the production frontier and calculate input redundancy and output gap rates, and evaluate low-carbon economic development. Through the establishment and application of these two models, the research achieves carbon emission prediction and low-carbon economic analysis, validating the feasibility of the research methodology. The results show that the composite model can effectively predict carbon emissions, with a Mean Absolute Percentage Error (MAPE) below 3.5% and Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) below 200 tons, demonstrating the feasibility and accuracy of the model. The research on low-carbon economic development in S Province based on the DEA method reveals the need for energy structure adjustment, clean and renewable energy promotion, control of carbon emissions, and optimization of industrial structure with a focus on developing the tertiary industry. Therefore, the use of artificial intelligence and big data analysis can provide more precise insights into the trends and patterns of low-carbon economic development, as well as more effective predictions of future energy demand and resource supply, offering high practical value and scientific significance."}97{"work_id": 4411541921, "title": "From Efficiency Gains to Rebound Effects: The Problem of Jevons' Paradox in AI's Polarized Environmental Debate", "abstract": "As the climate crisis deepens, artificial intelligence (AI) has emerged as a contested force: some champion its potential to advance renewable energy, materials discovery, and large-scale emissions monitoring, while others underscore its growing carbon footprint, water consumption, and material resource demands. Much of this debate has concentrated on direct impacts -- energy and water usage in data centers, e-waste from frequent hardware upgrades -- without addressing the significant indirect effects. This paper examines how the problem of Jevons' Paradox applies to AI, whereby efficiency gains may paradoxically spur increased consumption. We argue that understanding these second-order impacts requires an interdisciplinary approach, combining lifecycle assessments with socio-economic analyses. Rebound effects undermine the assumption that improved technical efficiency alone will ensure net reductions in environmental harm. Instead, the trajectory of AI's impact also hinges on business incentives and market logics, governance and policymaking, and broader social and cultural norms. We contend that a narrow focus on direct emissions misrepresents AI's true climate footprint, limiting the scope for meaningful interventions. We conclude with recommendations that address rebound effects and challenge the market-driven imperatives fueling uncontrolled AI growth. By broadening the analysis to include both direct and indirect consequences, we aim to inform a more comprehensive, evidence-based dialogue on AI's role in the climate crisis."}98{"work_id": 4362707075, "title": "Making AI Less \"Thirsty\": Uncovering and Addressing the Secret Water Footprint of AI Models", "abstract": "The growing carbon footprint of artificial intelligence (AI) has been undergoing public scrutiny. Nonetheless, the equally important water (withdrawal and consumption) footprint of AI has largely remained under the radar. For example, training the GPT-3 language model in Microsoft's state-of-the-art U.S. data centers can directly evaporate 700,000 liters of clean freshwater, but such information has been kept a secret. More critically, the global AI demand is projected to account for 4.2-6.6 billion cubic meters of water withdrawal in 2027, which is more than the total annual water withdrawal of 4-6 Denmark or half of the United Kingdom. This is concerning, as freshwater scarcity has become one of the most pressing challenges. To respond to the global water challenges, AI can, and also must, take social responsibility and lead by example by addressing its own water footprint. In this paper, we provide a principled methodology to estimate the water footprint of AI, and also discuss the unique spatial-temporal diversities of AI's runtime water efficiency. Finally, we highlight the necessity of holistically addressing water footprint along with carbon footprint to enable truly sustainable AI."}99{"work_id": 4360605906, "title": "33.1 A 16nm 32Mb Embedded STT-MRAM with a 6ns Read-Access Time, a 1M-Cycle Write Endurance, 20-Year Retention at 150\u00b0C and MTJ-OTP Solutions for Magnetic Immunity", "abstract": "Embedded non-volatile memory (eNVM) is an essential element for microcontrollers (MCUs) used in automotive applications. As the automotive market transitions to greater electrification and autonomy, we are seeing MCU growth in the car: including integration to simplify system design, an electrical and electronic (E/E) architectural evolution to domain/zone control, and over-the-air (OTA) updates beyond 128Mb eNVM densities. To support this transition technology nodes are migrating from 55/40nm to 28/16nm. In addition, traditional charge-based embedded Flash will be replaced by Back-end-of-line (BEOL) memories: such as STT-MRAM [1]\u2013[3], PCRAM [4], and RRAM. Of these candidates, STT-MRAM is the most promising solution for automotive applications due to its high-temperature data retention, high write endurance, and fast write speed. However, STT-MRAM still faces several challenges from the inherent properties of magnetic tunnel junctions (MTJs) and the side effects of process integration: such as array-level variability and magnetic-field interference (MFI). In this work, several design solutions are proposed to overcome these challenges: (1) a novel merged-local-reference scheme is used to overcome the array-edge effect on the MTJs; (2) write bias segment trimming is used to mitigate the near-far effect for better write endurance; (3) an MTJ-based one-time programmable (OTP) is used to preserve critical data during the wafer-level chip-scale packaging (WLCSP) process (360\u00b0C, 3hr), and; (4) a novel sensing-reference scheme is used so that this MTJ-OTP is immune to external-magnetic-field interference. A 32-Mb STT-MRAM test chip based on these proposed solutions is successfully fabricated in a 16nm FinFET CMOS process, Measured results confirm its excellent performance and manufacturability for next generation automotive MCU application."}100{"work_id": 2799210008, "title": "Memristor Crossbars with 4.5 Terabits-per-Inch-Square Density and Two Nanometer Dimension", "abstract": "Memristor is a promising building block for the next generation nonvolatile random access memory and bio-inspired computing systems. Organizing memristors into high density crossbar arrays, although challenging, is critical to meet the ever-growing high capacity and low energy demands of these applications especially in the big data era. Here, we construct memristor crossbars with a single-layer density up to 4.5 terabits per inch square, an order of magnitude denser than the state- of-the-art 64-layer triple level cell NAND flash technology. The memristors in the crossbars are 2 $\\times$ 2 nm$^2$ in size, capable of switching with tens of nano ampere electric current. The densely packed memristor crossbars of extremely small working devices provides a power-efficient solution for high density information storage and processing."}101{"work_id": 4210357113, "title": "Opportunities for neuromorphic computing algorithms and applications", "abstract": null}102{"work_id": 4408612646, "title": "Microsatellite-based real-time quantum key distribution", "abstract": null}103{"work_id": 7149961640, "title": "A discrete memristive cyclic Hopfield neural network with multi-cavity-like attractors and application in secure communication", "abstract": null}104{"work_id": 3181035463, "title": "{ZNS}: Avoiding the Block Interface Tax for Flash-based SSDs", "abstract": null}105{"work_id": 7117963061, "title": "Toward sustainable data center operation: A review on existing infrastructures, integrated smart energy management frameworks, and future perspectives", "abstract": null}106{"work_id": 1983572666, "title": "The Art of Multiprocessor Programming", "abstract": null}107{"work_id": 4414574650, "title": "Field and delay effects on synchronization in a multilayer memristive neuronal network with a hub", "abstract": null}108{"work_id": 4413052649, "title": "An attractor-controllable memristive Hopfield neural network and its application on voice encryption", "abstract": null}109{"work_id": 2482126025, "title": "Characterizing quantum supremacy in near-term devices", "abstract": null}110{"work_id": 7141052859, "title": "A dual-neuron second-order memristor-based Hopfield neural network with controllable extreme multistability and its application in image encryption", "abstract": null}111{"work_id": 2798434869, "title": "Supervised learning with quantum-enhanced feature spaces", "abstract": null}112{"work_id": 2803823791, "title": "A review of thermal management and innovative cooling strategies for data center", "abstract": null}113{"work_id": 1653864426, "title": "Inside Solid State Drives (SSDs)", "abstract": null}114{"work_id": 4409283612, "title": "Universal photonic artificial intelligence acceleration", "abstract": null}115{"work_id": 2944119451, "title": "All-optical spiking neurosynaptic networks with self-learning capabilities", "abstract": null}116{"work_id": 2892611063, "title": "Improving Big Data Centers Energy Efficiency: Traffic Based Model and Method", "abstract": null}117{"work_id": 4408847956, "title": "Barren plateaus in variational quantum computing", "abstract": null}118{"work_id": 7130682590, "title": "Extreme multistability in discrete memristive neuron maps and implications for dual-field applications", "abstract": null}119{"work_id": 7140216027, "title": "Discrete bidirectional memristive neural network-based hyperchaotic system and its FPGA implementation", "abstract": null}120{"work_id": 4403931219, "title": "BPF Instruction Set Architecture (ISA)", "abstract": null}121{"work_id": 7162190207, "title": "AI-driven mental health therapy: A comparative study of emotional clarity and stigma reduction across journaling, human-guided therapy, and AI-guided therapy", "abstract": null}122{"work_id": 4391530352, "title": "Data centers cooling: A critical review of techniques, challenges, and energy saving solutions", "abstract": null}123{"work_id": 3111721177, "title": "A Modern Primer on Processing in Memory", "abstract": null}124{"work_id": 7134974875, "title": "Neuromorphic Photonic Computing using Microring Resonator Arrays for Real-Time Edge AI Processing", "abstract": null}125{"work_id": 3003967059, "title": "Parallel convolutional processing using an integrated photonic tensor core", "abstract": null}126