A5090386129- Neural Networks and Reservoir Computing
- Optical Network Technologies
- Photonic and Optical Devices
- Quantum Computing Algorithms and Architecture
- Advanced Memory and Neural Computing
- Quantum Information and Cryptography
- Advanced Neural Network Applications
- VLSI and FPGA Design Techniques
- Advancements in Semiconductor Devices and Circuit Design
- Low-power high-performance VLSI design
- VLSI and Analog Circuit Testing
- Domain Adaptation and Few-Shot Learning
- Parallel Computing and Optimization Techniques
- Advanced Database Systems and Queries
- Semantic Web and Ontologies
- Quantum and electron transport phenomena
- Energy Load and Power Forecasting
- Advanced Photonic Communication Systems
- Robotics and Sensor-Based Localization
- Electric Power System Optimization
- Energy, Environment, Economic Growth
- Formal Methods in Verification
- Advancements in Photolithography Techniques
- Semiconductor Lasers and Optical Devices
- Data Management and Algorithms
Arizona State University
2023-2025
Southeast University
2025
Zhongda Hospital Southeast University
2025
State Key Laboratory of Digital Medical Engineering
2025
Hohai University
2020-2024
The University of Texas at Austin
2019-2024
Guangxi University for Nationalities
2024
Nanjing Medical University
2018-2024
Beijing Anzhen Hospital
2023
Capital Medical University
2023
Vision Transformers (ViTs) have emerged with superior performance on computer vision tasks compared to the convolutional neural network (CNN)-based models. However, ViTs mainly designed for image classification will generate single-scale low-resolution representations, which makes dense prediction such as semantic segmentation challenging ViTs. Therefore, we propose HRViT, enhances learn semantically-rich and spatially-precise multi-scale representations by integrating high-resolution...
Quantum noise is the key challenge in Noisy Intermediate-Scale (NISQ) computers. Previous work for mitigating has primarily focused on gate-level or pulse-level noise-adaptive compilation. However, limited research explored a higher level of optimization by making quantum circuits themselves resilient to noise.In this paper, we propose QuantumNAS, comprehensive framework co-search variational circuit and qubit mapping. Variational are promising approach constructing neural networks machine...
Abstract The past two decades have witnessed the stagnation of clock speed microprocessors followed by recent faltering Moore’s law as nanofabrication technology approaches its unavoidable physical limit. Vigorous efforts from various research areas been made to develop power-efficient and ultrafast computing machines in this post-Moore’s era. With unique capacity integrate complex electro-optic circuits on a single chip, integrated photonics has revolutionized interconnects shown striking...
Placement for very large-scale integrated (VLSI) circuits is one of the most important steps design closure. We propose a novel GPU-accelerated placement framework DREAMPlace, by casting analytical problem equivalently to training neural network. Implemented on top widely adopted deep learning toolkit PyTorch, with customized key kernels wirelength and density computations, DREAMPlace can achieve around 40× speedup in global without quality degradation compared state-of-the-art multithreaded...
Fast and accurate pre-routing timing prediction is essential for timing-driven placement since repetitive routing static analysis (STA) iterations are expensive unacceptable. Prior work on aims at estimating net delay slew, lacking the ability to model global metrics. In this work, we present a engine inspired graph neural network (GNN) predict arrival time slack endpoints. We further leverage edge delays as local auxiliary tasks facilitate training with increased performance. Experimental...
The optical neural network (ONN) is a promising hardware platform for next-generation neurocomputing due to its high parallelism, low latency, and energy consumption. Previous ONN architectures are mainly designed general matrix multiplication (GEMM), leading unnecessarily large area cost control complexity. Here, we move beyond classical GEMM-based ONNs propose an subspace (OSNN) architecture, which trades the universality of weight representation lower component usage, cost, We devise...
Parameterized Quantum Circuits (PQC) are promising towards quantum advantage on near-term hardware. However, due to the large noises (errors), performance of PQC models has a severe degradation real devices. Take Neural Network (QNN) as an example, accuracy gap between noise-free simulation and noisy results IBMQ-Yorktown for MNIST-4 classification is over 60%. Existing noise mitigation methods general ones without leveraging unique characteristics PQC; other hand, existing work does not...
Optical neural networks (ONNs) are promising hardware platforms for next-generation neuromorphic computing due to their high parallelism, low latency, and energy consumption. However, previous integrated photonic tensor cores (PTCs) consume numerous single-operand optical modulators signal weight encoding, leading large area costs propagation loss implement operations. This work proposes a scalable efficient dot-product engine based on customized multi-operand devices, namely neuron (MOON)....
Electronic–photonic computing systems offer immense potential in energy-efficient artificial intelligence (AI) acceleration tasks due to the superior speed and efficiency of optics, especially for real-time, low-energy deep neural network inference on resource-restricted edge platforms. However, current optical accelerators based foundry-available devices conventional system architecture still encounter a performance gap compared highly customized electronic counterparts. To bridge lack...
Placement is an important step in modern verylarge-scale integrated (VLSI) designs. Detailed placement a refining procedure intensively called throughout the design flow, thus its efficiency has vital impact on closure. However, since most detailed techniques are inherently greedy and sequential, they generally difficult to parallelize. In this article, we present concurrent framework, ABCDPlace, exploiting multithreading graphic processing unit (GPU) acceleration. We propose batch-based...
As a promising neuromorphic framework, the optical neural network (ONN) demonstrates ultra-high inference speed with low energy consumption. However, previous ONN architectures have high area overhead which limits their practicality. In this paper, we propose an area-efficient architecture based on structured networks, leveraging fast Fourier transform for efficient computation. A two-phase software training flow pruning is proposed to further reduce component utilization. Experimental...
With the rapid growth of photovoltaic (PV) power in recent years, stability system operation, performance contingency analysis as well quality grid are threatened by inherent uncertainty and fluctuation PV output. It is necessary to have knowledge output characteristics for reliable dispatching. Day-ahead forecasting an effective support achieving optimal Probabilistic can describe that difficult depict deterministic forecasting, results more comprehensive. An ensemble nonparametric...
Parameterized Quantum Circuits (PQC) are drawing increasing research interest thanks to its potential achieve quantum advantages on near-term Noisy Intermediate Scale (NISQ) hardware. In order scalable PQC learning, the training process needs be offloaded real machines instead of using exponential-cost classical simulators. One common approach obtain gradients is parameter shift whose cost scales linearly with number qubits. We present QOC, first experimental demonstration practical on-chip...
As deep learning models and datasets rapidly scale up, model training is extremely time-consuming resource-costly. Instead of on the entire dataset, with a small synthetic dataset becomes an efficient solution. Extensive research has been explored in direction condensation, among which gradient matching achieves state-of-the-art performance. The method directly targets dynamics by when original datasets. However, there are limited investigations into principle effectiveness this method. In...
The wide adoption and significant computing resource cost of attention-based transformers, e.g., Vision Transformers large language models, have driven the demand for efficient hardware accelerators. While electronic accelerators been commonly used, there is a growing interest in exploring photonics as an alternative technology due to its high energy efficiency ultra-fast processing speed. Photonic demonstrated promising results convolutional neural networks (CNNs) workloads, which...
Despite tremendous efforts in analog layout automation, little adoption has been demonstrated practical design flows. Traditional synthesis tools use various heuristic constraints to prune the space ensure post performance. However, these approaches provide limited guarantee and poor generalizability due a lack of model mapping properties circuit In this paper, we attempt shorten gap performance modeling for circuits with quantitative statistical approach. We leverage state-of-the-art...
ABSTRACT Transjugular intrahepatic portosystemic shunt (TIPS) is a widely used surgery for portal hypertension. In clinical practice, the diameter of stent forming usually selected empirically, which will influence postoperative pressure. Clinical studies found that inappropriate pressure after TIPS responsible poor prognosis; however, there no scheme to predict Therefore, this study aims develop computational model applied ahead surgery. For purpose, patient‐specific 0‐3‐D multi‐scale...
Optical simulation plays an important role in photonic hardware design flow. The finite-difference time-domain (FDTD) method is widely adopted to solve Maxwell equations. However, FDTD known for its prohibitive runtime cost as it iteratively solves equations and takes minutes hours simulate a single device. Recently, AI has been applied realize orders-of-magnitude speedup partial differential equation solving. AI-based solvers devices have not clearly formulated. Directly applying...