A5019374620- Neural Networks and Reservoir Computing
- Advanced Memory and Neural Computing
- Optical Network Technologies
- Adversarial Robustness in Machine Learning
- Neural dynamics and brain function
- Photonic and Optical Devices
- Advanced Neural Network Applications
- VLSI and Analog Circuit Testing
- Liver Disease Diagnosis and Treatment
- Embedded Systems Design Techniques
- Pediatric Hepatobiliary Diseases and Treatments
- 3D Shape Modeling and Analysis
- Real-time simulation and control systems
- 3D Printing in Biomedical Research
- Anomaly Detection Techniques and Applications
- Drug Transport and Resistance Mechanisms
- Gallbladder and Bile Duct Disorders
- Physical Unclonable Functions (PUFs) and Hardware Security
- Advanced Image Fusion Techniques
- Privacy-Preserving Technologies in Data
- Photoreceptor and optogenetics research
- Pharmacological Effects of Natural Compounds
- Computer Graphics and Visualization Techniques
- Advanced Fiber Laser Technologies
- Cellular Mechanics and Interactions
Xidian University
2020-2024
Harbin Engineering University
2024
Xuzhou Medical College
2023
Hangzhou Dianzi University
2021-2022
University of Electronic Science and Technology of China
2011-2022
San Jose State University
2021-2022
Zhejiang University of Technology
2021
National University of Defense Technology
2021
University of Yamanashi
2021
Columbia University
2018-2020
Abstract Current liver fibrosis scoring by computer-assisted image analytics is not fully automated as it requires manual preprocessing (segmentation and feature extraction) typically based on domain knowledge in pathology. Deep learning-based algorithms can potentially classify these images without the need for through learning from a large dataset of images. We investigated performance classification models built using deep algorithm pre-trained multiple sources to score compared them...
We propose computing primitive for an all-optical spiking neural network (SNN) based on vertical-cavity surface-emitting lasers (VCSELs) supervised learning by using biologically plausible mechanisms. The spike-timing-dependent plasticity (STDP) model was established the dynamics of semiconductor optical amplifier (VCSOA) subject to dual-optical pulse injection. neuron-synapse self-consistent unified SNN developed, which enables reproducing essential neuron-like and STDP function. Optical...
Abstract The explosive growth of data and information has motivated various emerging non-von Neumann computational approaches in the More-than-Moore era. Photonics neuromorphic computing attracted lots attention due to fascinating advantages such as high speed, wide bandwidth, massive parallelism. Here, we offer a review on optical neural our research groups at device system levels. photonics neuron synapse plasticity are presented. In addition, introduce several architectures algorithms...
Photonic neuromorphic computing has emerged as a promising approach to building low-latency and energy-efficient non-von Neuman system. A photonic spiking neural network (PSNN) exploits brain-like spatiotemporal processing realize high-performance computing. However, the nonlinear computation of PSNN remains significant challenge. Here, we propose fabricate neuron chip based on an integrated Fabry–Perot laser with saturable absorber (FP-SA). The neuron-like dynamics including temporal...
We propose a simple hardware architecture for solving exclusive OR (XOR) tasks in single step by using photonic spiking neuron based on vertical-cavity surface-emitting lasers with an embedded saturable absorber (VCSEL-SA) subject to dual-polarized pulsed optical injection. model the inhibitory extending Yamada and spin-flip incorporate two polarization-resolved modes absorber. It is shown that, carefully adjusting temporal difference according window, XOR operation can be realized neuron,...
Neuromorphic photonic computing has emerged as a competitive paradigm to overcome the bottlenecks of von-Neumann architecture. Linear weighting and nonlinear spike activation are two fundamental functions spiking neural network (PSNN). However, they separately implemented with different materials devices, hindering large-scale integration PSNN. Here, we propose, fabricate experimentally demonstrate neuro-synaptic chip enabling simultaneous implementation linear based on distributed feedback...
We propose and experimentally demonstrate the generation of dual-channels chaos with time delay signature (TDS) concealment by introducing a phase-modulated Sagnac loop in mutually coupled semiconductor lasers (MCSL). Furthermore, we utilization to solve multi-armed bandit (MAB) problem reinforcement learning. The experimental results agree well numerical simulations. For purpose comparison, MCSL conventional is also considered. It found that TDS chaotic signals can be better concealed our...
According to the World Health Organization (WHO), wearing a face mask is one of most effective protections from airborne infectious diseases such as COVID-19. Since spread COVID-19, infected countries have been enforcing strict regulation for indoor businesses and public spaces. While requirement, position type should also be considered in order increase effectiveness masks, especially at specific locations. However, this makes it difficult conventional facial recognition technology identify...
We propose a fully-connected photonic spiking neural network (SNN) consisting of excitable vertical-cavity surface-emitting lasers with an embedded saturable absorber (VCSELs-SA) to implement spike sequence learning by supervised training. The spike-timing-dependent plasticity (STDP) is incorporated into classical remote method (ReSuMe) algorithm training SNN for the first time. computation model derived based on Yamada model. To optimize process, we further novel measure, so-called...
We propose a framework for hardware architecture and learning algorithm co-design of multi-layer photonic spiking neural network (SNN). The vertical-cavity surface-emitting laser with an embedded saturable absorber (VCSEL-SA) which contains two polarization-resolved modes is employed as neuron. connection between identical polarization considered the excitatory synapse, whereas orthogonal regarded inhibitory synapse. physical model neuron derived based on combination spin-flip Yamada model....
Although deep neural networks (DNNs) have been increasingly applied in industrial cyber physical systems (ICPSs), they are vulnerable to security attacks due the tight interaction between elements and elements. In this article, we aim protect core IP of DNNs, i.e., model weights, against attacks. Different from conventional approaches, a layerwise protection framework is proposed ensure confidentiality DNN weights during inference procedure such that quality maximized, while satisfying...
There is widespread evidence that increasing functional mass of brown adipose tissue (BAT) via browning white (WAT) could potentially counter obesity and diabetes. However, most current approaches focus on administration pharmacological compounds which expose patients to highly undesirable side effects. Here, we describe a simple direct tissue-grafting approach increase BAT through ex vivo subcutaneous WAT, followed by re-implantation into the host; this cell-therapy act synergistically with...
Organoids are becoming widespread in drug-screening technologies but have been used sparingly for cell therapy as current approaches producing self-organized clusters lack scalability or reproducibility size and cellular organization. We introduce a method of using hydrogels sacrificial scaffolds, which allow cells to form followed by gentle release, resulting highly reproducible multicellular structures on large scale. demonstrated this strategy endothelial mesenchymal stem self-organize...
Exposure to teratogenic chemicals during pregnancy may cause severe birth defects. Due high inter-species variation of drug responses as well financial and ethical burdens, despite the widely use in vivo animal tests, it's crucial develop highly predictive human pluripotent stem cell (hPSC)-based vitro assays identify potential teratogens. Previously we have shown that morphological disruption mesoendoderm patterns formed by geometrically-confined differentiation migration using hPSCs could...
To explore potential risk factors of isolated diastolic hypertension (IDH) among young and middle-aged Chinese.A community-based cross-sectional study was conducted 338 subjects, aged 25 years above, using random sampling technique. There were 68 cases IDH, 46 systolic (ISH), 89 (SDH), 135 subjects with normal blood pressure. Cases controls matched on sex by frequency matching. Demographic characteristics, pressure other relevant information collected.Compared controls, patients IDH ISH had...
We propose an inferencing framework of a hybrid-integrated photonic spiking neural network (PSNN) to perform pattern recognition tasks, where the linear computation is realized based on 4 × silicon Mach-Zehnder interferometer (MZI) array, and nonlinear performed by InP-based neuron array vertical-cavity surface-emitting lasers with embedded saturable absorber (VCSELs-SA). With modified Tempotron-like remote supervised method (ReSuMe) training algorithm, we realize two numbers "0-3" optical...
As an emerging threat to deep neural networks (DNNs), backdoor attacks have received increasing attentions due the challenges posed by lack of transparency inherent in DNNs. In this article, we develop efficient algorithm from interpretability DNNs defend against DNN models. To extract critical neurons, deploy sets control gates following neurons layers, and function a model can be interpreted as semantic sensitivities input samples. A identification approach, derived activation frequency...
Spiking neural networks (SNNs) utilize brain-like spatiotemporal spike encoding for simulating brain functions. Photonic SNN offers an ultrahigh speed and power efficiency platform implementing high-performance neuromorphic computing. Here, we proposed a multi-synaptic photonic SNN, combining the modified remote supervised learning with delay-weight co-training to achieve pattern classification. The impact of connections robustness network were investigated through numerical simulations. In...