A5041025969- Parallel Computing and Optimization Techniques
- Advanced Neural Network Applications
- Advanced Data Storage Technologies
- Cloud Computing and Resource Management
- Distributed and Parallel Computing Systems
- Data Visualization and Analytics
- Stochastic Gradient Optimization Techniques
- Computer Graphics and Visualization Techniques
- Brain Tumor Detection and Classification
- Domain Adaptation and Few-Shot Learning
- Vacuum and Plasma Arcs
- Machine Learning and ELM
- UAV Applications and Optimization
- Machine Learning and Algorithms
- Robotic Path Planning Algorithms
- Embedded Systems Design Techniques
- GNSS positioning and interference
- Adversarial Robustness in Machine Learning
- Autonomous Vehicle Technology and Safety
- Ferroelectric and Negative Capacitance Devices
- Indoor and Outdoor Localization Technologies
- Radio Wave Propagation Studies
- Electrostatic Discharge in Electronics
- Complex Systems and Decision Making
- Data Stream Mining Techniques
Williams (United States)
2020-2025
William & Mary
2020-2025
Dalian University of Technology
2022-2025
Baidu (China)
2022-2025
Northwest A&F University
2024-2025
Shanghai Electric (China)
2025
Second Affiliated Hospital of Nanjing Medical University
2025
Ningbo University
2024
Beijing Jiaotong University
2024
Fujian University of Traditional Chinese Medicine
2023-2024
Deep neural networks have demonstrated state-of-the-art performance in a variety of real-world applications. In order to obtain gains, these grown larger and deeper, containing millions or even billions parameters over thousand layers. The tradeoff is that large architectures require an enormous amount memory, storage, computation, thus limiting their usability. Inspired by the recent tensor ring factorization, we introduce Tensor Ring Networks (TR-Nets), which significantly compress both...
Graphics Processing Units (GPUs) can easily outperform CPUs in processing large-scale data parallel workloads, but are considered weak serialized tasks and communicating with other devices. Pursuing a CPU-GPU collaborative computing model which takes advantage of both devices could provide an important breakthrough realizing the full performance potential heterogeneous computing. In recent years platform vendors runtime systems have added new features such as unified memory space dynamic...
In response to COVID-19, a vast number of visualizations have been created communicate information the public. Information exposure in public health crisis can impact people's attitudes towards and responses risks, ultimately trajectory pandemic. As such, there is need for work that documents, organizes, investigates what COVID-19 presented We address this gap through an analysis 668 visualizations. present our findings conceptual framework derived from analysis, examines who, (uses) data,...
The integration of AI assistants, especially through the development Large Language Models (LLMs), into computer science education has sparked significant debate, highlighting both their potential to augment student learning and risks associated with misuse. An emerging body work looked using LLMs in education, primarily focusing on evaluating performance existing models or conducting short-term human subject studies. However, very little examined impacts LLM-powered assistants students...
The rapidly growing popularity and scale of data-parallel workloads demand a corresponding increase in raw computational power Graphics Processing Units (GPUs). As single-GPU platforms struggle to satisfy these performance demands, multi-GPU have started dominate the high-performance computing world. advent such systems raises number design challenges, including GPU microarchitecture, interconnect fabric, runtime libraries, associated programming models. research community currently lacks...
Moore's Law and Dennard Scaling have guided the semiconductor industry for past few decades. Recently, both laws faced validity challenges as transistor sizes approach practical limits of physics. We are interested in testing these reflect on reasons responsible. In this work, we collect data more than 4000 publicly-available CPU GPU products. find that scaling remains critical keeping valid. However, architectural solutions become increasingly important will play a larger role future....
Cross-domain weakly supervised object detection (CD-WSOD) aims to adapt the model a novel target domain with easily acquired image-level annotations. How align source and domains is critical CDWSOD accuracy. Existing methods usually focus on partial components for alignment. In contrast, this paper considers that all are important proposes Holistic Hier-archical Feature Alignment (H <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> FA)...
During crises like COVID-19, individuals are inundated with conflicting and time-sensitive information that drives a need for rapid assessment of the trustworthiness reliability sources platforms. This parallels evolutions in infrastructures, ranging from social media to government data Distinct current literature, which presumes static relationship between presence or absence trust people's behaviors, our mixed-methods research focuses on situated trust, is shaped by information-seeking...
Understanding the impact of CYP2D6 metabolism on paroxetine, a widely used antidepressant, is essential for precision dosing.
High performance multi-GPU systems are widely used to accelerate training of deep neural networks (DNNs) by exploiting the inherently massive parallel nature process. Typically, DNNs in leverages a data-parallel model which DNN is replicated on every GPU, and each GPU performs Forward Propagation (FP), Backward (BP) and, Weight Update (WU). We analyze WU stage that composed collective communication (e.g., allReduce, broadcast), demands very efficient among GPUs avoid diminishing returns when...
During the COVID-19 pandemic, a number of data visualizations were created to inform public about rapidly evolving crisis. Data dashboards, form information dissemination used during have facilitated this process by visualizing statistics regarding cases over time. Prior work on has primarily focused design and evaluation specific visualization systems from technology-centered perspectives. However, little is known what occurs behind scenes creation processes, given complex sociotechnical...
This paper presents a DETR-based method for cross-domain weakly supervised object detection (CDWSOD), aiming at adapting the detector from source to target domain through weak supervision. We think DETR has strong potential CDWSOD due an insight: encoder and decoder in are both based on attention mechanism thus capable of aggregating semantics across entire image. The aggregation results, i.e., image-level predictions, can naturally exploit supervision alignment. Such motivated, we propose...
Unexpected disasters, both naturally occurring and those caused through human actions, result in severe damage to communication infrastructure. Additionally, such events are accompanied by sharp spikes the usage of commercially licensed spectrum, when affected victims tragedy attempt transmit information about themselves capture high bandwidth data form pictures videos. We envisage cognitive radio as a candidate solution situations, where devices can identify alternate frequency bands,...
Heterogeneous systems, that marry CPUs and GPUs together in a range of configurations, are quickly becoming the design paradigm for today's platforms because their impressive parallel processing capabilities. However, many existing heterogeneous GPU is only treated as an accelerator by CPU, working slave to CPU master. But recently we starting see introduction new class devices changes system runtime model, which enable accelerators be first-class computing devices. To support...
As transistor scaling becomes increasingly more difficult to achieve, the core count on a single GPU chip has also become extremely challenging. volume of data process in today's parallel workloads continues grow unbounded, we need find scalable solutions that can keep up with this increasing demand. To meet modern-day applications, multi-GPU systems offer promising path deliver high performance and large memory capacity. However, suffer from issues associated GPU-to-GPU communication...
Graph Neural Networks (GNNs) have emerged as a promising class of Machine Learning algorithms to train on non-euclidean data. GNNs are widely used in recommender systems, drug discovery, text understanding, and traffic forecasting. Due the energy efficiency high-performance capabilities GPUs, GPUs natural choice for accelerating training GNNs. Thus, we want better understand architectural system-level implications GPUs. Presently, there is no benchmark suite available designed study GNN...
The dynamic inference, which adaptively allocates computational budgets for different samples, is a prevalent approach achieving efficient action recognition. Current studies primarily focus on data-efficient regime that reduces spatial or temporal redundancy, their combination, by selecting partial video data, such as clips, frames, patches. However, these approaches often utilize fixed and computationally expensive networks. From perspective, this article introduces novel model-efficient...
We investigate a family of approximate multi-step proximal point methods, framed as implicit linear discretizations gradient flow. The resulting methods are with similar computational cost in each update the method. explore several optimization where applying an multistep points method results improved convergence behavior. also include analysis for proposed problem settings: quadratic problems, general problems that strongly or weakly (non)convex, and accelerated alternating projections.
The conduction delay time is one of the main technical parameters surface flashover triggered vacuum switch (STVS). Detailed research and optimization characteristics are necessary for development high-performance STVS. In this paper, performance STVS with different electrode structures operating investigated in detail. Based on analysis process, factors affecting discussed. experimental platform built by using a detachable chamber, compared under parameters. This analyzes discusses reasons...
Recent advancements in large language models (LLMs) have spurred growing interest automatic theorem proving using Lean4, where effective tree search methods are crucial for navigating proof spaces. While the existing approaches primarily rely on value functions and Monte Carlo Tree Search (MCTS), potential of simpler like Best-First (BFS) remains underexplored. This paper investigates whether BFS can achieve competitive performance large-scale tasks. We present \texttt{BFS-Prover}, a...
Generative AI (GenAI) has brought opportunities and challenges for higher education as it integrates into teaching learning environments. As instructors navigate this new landscape, understanding their engagement with attitudes toward GenAI is crucial. We surveyed 178 from a single U.S. university to examine current practices, perceptions, trust, distrust of in March 2024. While most reported moderate high familiarity GenAI-related concepts, actual use tools direct instructional tasks...