A5026654173- Advanced Memory and Neural Computing
- Neural dynamics and brain function
- Neural Networks and Reservoir Computing
- Neural Networks and Applications
- Distributed Control Multi-Agent Systems
- Neuroscience and Neural Engineering
- Video Surveillance and Tracking Methods
- Neural Networks Stability and Synchronization
- Adaptive Control of Nonlinear Systems
- Human Pose and Action Recognition
- EEG and Brain-Computer Interfaces
- Anomaly Detection Techniques and Applications
- Chemical and Physical Properties in Aqueous Solutions
- Face recognition and analysis
- Network Traffic and Congestion Control
- Advanced Neural Network Applications
- Free Radicals and Antioxidants
- Network Security and Intrusion Detection
- IoT and GPS-based Vehicle Safety Systems
- Digital Media Forensic Detection
- Gait Recognition and Analysis
- ECG Monitoring and Analysis
- Emotion and Mood Recognition
- Photochemistry and Electron Transfer Studies
- Industrial Technology and Control Systems
University of Macau
2022-2025
Anhui University
2023-2025
Institute of Microelectronics
2022-2025
Analog Devices (United States)
2025
Hong Kong Polytechnic University
2023
Ministry of Education of the People's Republic of China
2023
City University of Macau
2023
Zhuhai Institute of Advanced Technology
2023
Fuzhou University
2019-2022
Minjiang University
2022
Person re-identification (Re-ID) is to retrieve a particular person captured by different cameras, which of great significance for security surveillance and pedestrian behavior analysis. However, due the large intra-class variation across e.g., occlusions, illuminations, viewpoints, poses, Re-ID still challenging task in field computer vision. In this paper, attack issues concerning with variation, we propose coarse-to-fine framework incorporation auxiliary-domain classification (ADC)...
The biologically discovered intrinsic plasticity (IP) learning rule, which changes the excitability of an individual neuron by adaptively turning firing threshold, has been shown to be crucial for efficient information processing. However, this rule needs extra time updating operations at each step, causing energy consumption and reducing computational efficiency. event-driven or spike-based coding strategy spiking neural networks (SNNs), i.e., neurons will only active if driven continuous...
Spiking neural networks (SNNs) have captivated the attention worldwide owing to their compelling advantages in low power consumption, high biological plausibility, and strong robustness. However, intrinsic latency associated with SNNs during inference poses a significant challenge, impeding further development application. This is caused by need for spiking neurons collect electrical stimuli generate spikes only when membrane potential exceeds firing threshold. Considering threshold plays...
Multi-label pedestrian attribute recognition (PAR) involves assigning multiple attributes to images captured by video surveillance cameras. Despite its importance, learning robust attribute-related features for PAR remains a challenge due the large intra-attribute variations in image space. These variations, which stem from changes poses, illumination conditions, and background noise, make extracted susceptible irrelevant information or noise interference. Existing methods rely on body prior...
In this paper, a novel high-accuracy and robust computing framework for time series classification tasks is presented. The consists of feature extraction module module, where the implemented by reservoir method spiking neurons, result obtained state-of-the-art analog convolutional neural networks (CNNs). original input first converted to multi-channel spike streams, then fed into layer produce intermediate output, subsequently, output transformed 2D mapping image, deep CNN model applied...
A magnetic field enhanced the catalytic activity of Fe + HZSM-5, preventing deactivation zeolite and facilitating BTEX production.
Long-term (also called Clothing-Change) person re-identification (CC-reID) aims at confirming the identity of pedestrians captured diverse locations and/or times. Current CC-reID methods heavily rely on ID features learned by CNN architecture. However, with limited receptive fields, is hard to effectively explore some unique but discriminative (e.g., hair style, tattoo and accessories) from small body regions. Compared CNN, Transformer has certain merits in exploring more ID-unique <sup...
Nowadays, more and online content providers are offering multiple types of data services. To provide users with a better service experience, Quality Experience (QoE) has been widely used in the delivery quality measurement network How to accurately measure QoE score for all services become meaningful but difficult problem. solve this problem, we proposed unified scoring framework that measures user experience almost The first uses machine learning model (random forest) classify services,...
We propose two simple and effective spiking neuron models to improve the response time of conventional neural network. The proposed adaptively tune presynaptic input current depending on received from its presynapses subsequent firing events. analyze derive activity homeostatic convergence models. experimentally verify compare MNIST handwritten digits FashionMNIST classification tasks. show that significantly increase speed signal. Experiment codes are available at <uri...
The Conceptor network is a newly proposed reservoir computing (RC) model, which outperforms traditional classifiers, can fail to model new classes of data for supervised learning task. However, the structure design single, involving just random network, has strong coupling between nodes and limits ability. This study focused on topology problem, we propose complex Conceptor-based phase space reconstruction time series. Several dynamical systems were chosen build networks using algorithm....
Echo state networks (ESNs) with multi-clustered reservoir topology perform better in computing and robustness than those random topology. However, these ESNs have a complex topology, which leads to difficulties generation. This study focuses on the generation problem when ESN is used environments sufficient priori data available. Accordingly, data-driven multi-cluster algorithm proposed. The proposed are evaluate reservoirs by calculating precision standard deviation of ESNs. produced using...
Intrinsic plasticity (IP) is an unsupervised, self-adaptive, local learning rule that was first found in biological nerve cells, and has been shown to be able maximize neuronal information transmission entropy. In this article, we propose a soft-reset leaky integrate-and-fire (LIF) model, spiking neuron model based on widely used LIF neurons, with new IP optimizes the membrane potential state exponentially distributed. Previous studies have generally such as expected firing rate target...
In this article, we present an echo state network (ESN)-based tracking control approach for a class of affine nonlinear systems. Different from the most existing neural-network (NN)-based methods that are focused on feedforward NN, proposed method adopts bioinspired recurrent NN fusing with multiple cluster and intrinsic plasticity (IP) to deal modeling uncertainties coupling nonlinearities in The key features work can be summarized as follows: 1) is built upon ESN embedded multiclustered...