A5054102866- Radiation Effects in Electronics
- Advanced Neural Network Applications
- Semiconductor materials and devices
- Advancements in Semiconductor Devices and Circuit Design
- Nuclear physics research studies
- Human Pose and Action Recognition
- Robotics and Sensor-Based Localization
- Nuclear Physics and Applications
- Radiation Detection and Scintillator Technologies
- Topic Modeling
- Radiation Therapy and Dosimetry
- Nuclear reactor physics and engineering
- 3D Shape Modeling and Analysis
- VLSI and Analog Circuit Testing
- Integrated Circuits and Semiconductor Failure Analysis
- Advanced Vision and Imaging
- Advanced Graph Neural Networks
- Neural Networks and Applications
- Quantum, superfluid, helium dynamics
- Silicon Carbide Semiconductor Technologies
- Gamma-ray bursts and supernovae
- Pharmacological Effects and Toxicity Studies
- Mathematical Biology Tumor Growth
- Cold Atom Physics and Bose-Einstein Condensates
- Hybrid Renewable Energy Systems
Columbia University
2025
Yantai University
2022-2025
Central South University
2024-2025
South China University of Technology
2024
Harbin Institute of Technology
2024
Yale University
2022-2024
Zhejiang Gongshang University
2024
Massachusetts Institute of Technology
1982-2024
Unitec Institute of Technology
2023-2024
Chongqing Medical University
2023-2024
Multi-sensor fusion is essential for an accurate and reliable autonomous driving system. Recent approaches are based on point-level fusion: augmenting the LiDAR point cloud with camera features. However, camera-to-LiDAR projection throws away semantic density of features, hindering effectiveness such methods, especially semantic-oriented tasks (such as 3D scene segmentation). In this paper, we propose BEVFusion, efficient generic multi-task multi-sensor framework. It unifies multi-modal...
New ground-based measurements of the cosmic-ray induced neutron flux and its energy distribution have been made at several locations across United States using an extended-energy Bonner sphere spectrometer. The data cover over twelve decades energy, from meV to GeV. An expression scale other has developed a fit altitude dependence our literature for geomagnetic solar-activity monitor rates. In addition, analytic is provided which fits spectrum above about 0.4 MeV. important estimating...
Deep neural networks (DNNs) have achieved unprecedented success in the field of artificial intelligence (AI), including computer vision, natural language processing, and speech recognition. However, their superior performance comes at considerable cost computational complexity, which greatly hinders applications many resource-constrained devices, such as mobile phones Internet Things (IoT) devices. Therefore, methods techniques that are able to lift efficiency bottleneck while preserving...
Large language models (LLMs) have shown excellent performance on various tasks, but the astronomical model size raises hardware barrier for serving (memory size) and slows down token generation bandwidth). In this paper, we propose Activation-aware Weight Quantization (AWQ), a hardware-friendly approach LLM low-bit weight-only quantization. Our method is based observation that weights are not equally important: protecting only 1% of salient can greatly reduce quantization error. We then to...
Deep learning on point clouds has received increased attention thanks to its wide applications in AR/VR and autonomous driving. These require low latency high accuracy provide real-time user experience ensure safety. Unlike conventional dense workloads, the sparse irregular nature of poses severe challenges running CNNs efficiently general-purpose hardware. Furthermore, existing acceleration techniques for 2D images do not translate 3D clouds. In this paper, we introduce TorchSparse, a...
High-resolution images enable neural networks to learn richer visual representations. However, this improved performance comes at the cost of growing computational complexity, hindering their usage in latency-sensitive applications. As not all pixels are equal, skipping computations for less-important regions offers a simple and effective measure reduce computation. This, however, is hard be translated into actual speedup CNNs since it breaks regularity dense convolution workload. In paper,...
We present LongLoRA, an efficient fine-tuning approach that extends the context sizes of pre-trained large language models (LLMs), with limited computation cost. Typically, training LLMs long is computationally expensive, requiring extensive hours and GPU resources. For example, on length 8192 needs 16x computational costs in self-attention layers as 2048. In this paper, we speed up extension two aspects. On one hand, although dense global attention needed during inference, model can be...
Deep learning on point clouds plays a vital role in wide range of applications such as autonomous driving and AR/VR. These interact with people real-time edge devices thus require low latency energy. Compared to projecting the cloud 2D space, directly processing 3D yields higher accuracy lower #MACs. However, extremely sparse nature poses challenges hardware acceleration. For example, we need explicitly determine nonzero outputs search for neighbors (mapping operation), which is unsupported...
Large language models (LLMs) have shown remarkable potential in processing long sequences, yet efficiently serving these long-context remains challenging due to the quadratic computational complexity of attention prefilling stage and large memory footprint KV cache decoding stage. To address issues, we introduce LServe, an efficient system that accelerates long-sequence LLM via hybrid sparse attention. This method unifies different hardware-friendly, structured sparsity patterns for both...
The key issues of cosmic-ray-induced soft-error rates, SER (also referred to as single-event upset, SEU, rates) in microelectronic devices are discussed from the viewpoint fundamental atomic and nuclear interactions between high-energy particles semiconductors. From sea level moderate altitudes, cosmic ray spectrum is dominated by three particle species: nucleons (protons neutrons), pions, muons. characteristic features reactions these with light elements reviewed. A major cause soft errors...
This anthology contains contributions from eleven different groups, each developing and/or applying Monte Carlo-based radiation transport tools to simulate a variety of effects that result energy transferred semiconductor material by single particle event. The topics span basic mechanisms for single-particle induced failures applied tasks like websites predict on-orbit event failure rates using Carlo tools.
In real applications, person re-identification (re-id) is an inherently domain adaptive computer vision task which often requires the model trained on a group of people to perform well unlabeled dataset consisting another pedestrians without supervised fine-tuning. Furthermore, there are typically large number classes (people) with small samples belonging each class. Based characteristics re-id and general assumptions related adaptation, we put forward novel algorithm for cross-dataset...
3D neural networks are widely used in real-world applications (e.g., AR/VR headsets, self-driving cars). They required to be fast and accurate; however, limited hardware resources on edge devices make these requirements rather challenging. Previous work processes data using either voxel-based or point-based networks, but both types of models not hardware-efficient due the large memory footprint random access. In this paper, we study deep learning from efficiency perspective. We first...
We describe SEMM-2, a new simulation system for the analysis of radiation-induced single event upsets which builds on initial SEMM tool. Developed current and future CMOS technologies, SEMM-2 improves generation radiation events. The atomic databases ion energy loss with transport through device materials are generalized. Enhancements nuclear collision include more accurate efficient methods generating elastic events thorough treatment inelastic processes. present illustrative simulations...
We present here the first systematic and global analysis of nucleon-induced reactions on A30 systems at incident energies 50 MeV to 1 GeV by means a cascade-statistical approach. Model simulations are tested against available data from nuclei such as C, N, O, Al, Si. With one set inputs provided here, most inclusive observables, double differential spectra light ions heavy residual dominant channels, can be well reproduced. The model yields absolute cross sections calculations done without...
In this paper, we review the current status of single-event upsets caused by alpha-particles in IBM circuits and technology. While both cosmic radiation can induce upsets, alpha-particle-induced upset rate has become an increasingly important issue because are no longer limited to memory circuits. Latch have highly sensitive alpha-particles. The latch is one most critical issues for microprocessors requiring high performance reliability.
This article concerns the fully parabolic pursuit-prey chemotaxis system $$\displaylines{ u_t=\Delta u-\chi\nabla\cdot\left(u\nabla w\right) +u\left(\lambda_1-\mu_1 u^{r_1-1}+av\right), \quad x\in\Omega,\; t>0,\cr v_t=\Delta v+\xi\nabla\cdot\left(v\nabla z\right)+v\left(\lambda_2-\mu_2 v^{r_2-1}-bu\right),\quad w_t=\Delta w-w+v,\quad t>0,\\ z_t=\Delta z-z+u, t>0, }$$ in a bounded domain \(\Omega\subset\mathbb{R}^{N}\) \((N\geq1)\) with homogeneous Neumann boundary conditions, where...
Schizophrenia is a multifaceted mental disorder affecting approximately 1% of the global population, significantly disrupting cognitive function, emotion, and behaviour. Antipsychotic medications are primary treatment, targeting neurotransmitter regulation to alleviate symptoms. However, long-term use often leads drug tolerance resistance, diminishing treatment effectiveness potentially exacerbating patient's condition. This article examines mechanisms contributing factors associated with...