A5063174730- Advanced Memory and Neural Computing
- Microbial Natural Products and Biosynthesis
- Computational Drug Discovery Methods
- Photoreceptor and optogenetics research
- Synthesis and biological activity
- Neural Networks and Applications
- Robotics and Sensor-Based Localization
- vaccines and immunoinformatics approaches
- Domain Adaptation and Few-Shot Learning
- Machine Learning and Algorithms
- Robotics and Automated Systems
- Neural dynamics and brain function
- Multimodal Machine Learning Applications
- Ferroelectric and Negative Capacitance Devices
- Neural Networks and Reservoir Computing
Zhejiang University
2020-2024
Zhejiang Lab
2023
University of Illinois Urbana-Champaign
2022
ORCID
2020
Spiking Neural Networks (SNNs) are emerging as a brain-inspired alternative to traditional Artificial (ANNs), prized for their potential energy efficiency on neuromorphic hardware. Despite this, SNNs often suffer from accuracy degradation compared ANNs and face deployment challenges due fixed inference timesteps, which require retraining adjustments, limiting operational flexibility. To address these issues, our work considers the spatio-temporal property inherent in SNNs, proposes novel...
Machine-learning-based scoring functions (MLSFs) have gained attention for their potential to improve accuracy in binding affinity prediction and structure-based virtual screening (SBVS) compared classical SFs. Developing accurate MLSFs SBVS requires a large unbiased dataset that includes structurally diverse actives decoys. Unfortunately, most datasets suffer from hidden biases data insufficiency. Here, we developed topology-based conformation-based decoys database (ToCoDDB). The biological...
Spiking Neural Networks (SNNs) are considered more biologically realistic and power-efficient as they imitate the fundamental mechanism of human brain. Backpropagation (BP) based SNN learning algorithms that utilize deep frameworks have achieved good performance. However, those BP-based partially ignore bio-interpretability. In modeling spike activity for biological plausible SNNs, we examine three properties: multiplicity, adaptability, plasticity (MAP). Regarding propose a Multiple-Spike...
Recent insights have revealed that rate-coding is a primary form of information representation captured by surrogate-gradient-based Backpropagation Through Time (BPTT) in training deep Spiking Neural Networks (SNNs). Motivated these findings, we propose rate-based backpropagation, strategy specifically designed to exploit representations reduce the complexity BPTT. Our method minimizes reliance on detailed temporal derivatives focusing averaged dynamics, streamlining computational graph...
Semi-supervised learning (SSL) is an efficient framework that can train models with both labeled and unlabeled data, but may generate ambiguous non-distinguishable representations when lacking adequate samples. With human-in-the-loop, active iteratively select informative samples for labeling training to improve the performance in SSL framework. However, most existing approaches rely on pre-trained features, which not suitable end-to-end learning. To deal drawbacks of SSL, this paper, we...
Abstract Virtual screening (VS) based on molecular docking has emerged as one of the mainstream technologies drug discovery due to its low cost and high efficiency. However, scoring functions (SFs) implemented in most programs are not always accurate enough how improve their prediction accuracy is still a big challenge. Here, we propose an integrated platform called ASFP, web server for development customized SFs structure-based VS. There three main modules ASFP: 1) descriptor generation...