Wan Zhang

Wan Zhang

Auto-generated profile, based on publicly available information

Wan Zhang

Key Laboratory of Collaborative Intelligence Systems, Ministry of Education, School of Computer Science and Technology, Xidian University, Xian, China

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 May 2025 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
JWJianing Wang
Jianing Wang
ZHZheng Hua
Zheng Hua
WZWan Zhang
Wan Zhang
SHShengjia Hao
Shengjia Hao
YYYuqiong Yao
Yuqiong Yao
Maoguo GongMaoguo Gong
Maoguo Gong
Memory stability and learning flexibility in continual learning (CL) is a core challenge for cross-scene Hyperspectral Anomaly Detection (HAD) task. Biological neural networks can actively forget history knowledge that conflicts with the learning of new experiences by regulating learning-triggered synaptic expansion and synaptic convergence. Inspired by this phenomenon, we propose a novel Biologically-Inspired Continual Learning Generative Adversarial Network (CL-BioGAN) for augmenting continuous distribution fitting ability for cross-domain HAD task, where Continual Learning Bio-inspired Loss (CL-Bio Loss) and self-attention Generative Adversarial Network (BioGAN) are incorporated to realize forgetting history knowledge as well as involving replay strategy in the proposed BioGAN. Specifically, a novel Bio-Inspired Loss composed with an Active Forgetting Loss (AF Loss) and a CL loss is designed to realize parameters releasing and enhancing between new task and history tasks from a Bayesian perspective. Meanwhile, BioGAN loss with L2-Norm enhances self-attention (SA) to further balance the stability and flexibility for better fitting background distribution for open scenario HAD (OHAD) tasks. Experiment results underscore that the proposed CL-BioGAN can achieve more robust and satisfying accuracy for cross-domain HAD with fewer parameters and computation cost. This dual contribution not only elevates CL performance but also offers new insights into neural adaptation mechanisms in OHAD task.

12 August 2025 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
JWJianing Wang
Jianing Wang
ZHZheng Hua
Zheng Hua
WZWan Zhang
Wan Zhang
SHShengjia Hao
Shengjia Hao
YYYuqiong Yao
Yuqiong Yao
Maoguo GongMaoguo Gong
Maoguo Gong
In the above article [1], in (19), the original formulation has been revised and replaced with the following: \begin{equation*} \textbf {R}_{t}\gets \ \textbf {R}_{t-1}\cup \ k\left ({{\textbf B_{t}}}\right ),\textbf {R}_{0}=\phi . {5pt}\tag {19}\end{equation*}

16 February 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
JWJianing Wang
Jianing Wang
WZWan Zhang
Wan Zhang
YLYichen Liu
Yichen Liu
ZWZhendong Wang
Zhendong Wang
ZYZhuocheng Yu
Zhuocheng Yu
HHHaoyang He
Haoyang He
Remote sensing image change detection (RSI-CD) is a key task for remote sensing image interpretation. It faces inherent challenges such as sample scarcity, significant differences in multi-temporal data, and high model complexity of algorithm. Therefore, a novel parallel dual-temporal information fusion network with dual-stage contrastive learning framework (DC-PanNet) is proposed to deal with above-mentioned challenges. The DC-PanNet is constructed with parallel fusion dual-temporal information (PFDI) modules, which compact spectral and spatial features through parallelized point-wise and standard convolution, respectively. Meanwhile, aim at exploiting significant differences in multi-temporal phase datasets under limited labeled sample condition, DC-PanNet innovatively trained through the proposed dual-stage contrastive learning strategy: 1) In the pre-contrastive learning stage, DC-PanNet efficiently learns invariant representations by enhancing the correlation of cross-temporal phase features and enforcing the spatial alignment of the corresponding pixels to extract mutually information-rich features from unlabeled sample pairs. 2) In the post-contrastive learning stage, the limited labeling information is utilized to further enhance the detection capability by explicitly distinguishing changing region pairs through supervised contrast constraints. We conduct experiments on three homologous datasets: the Jiangsu, China, and US datasets, and three heterologous datasets: the California (multispectral-SAR), Sardinia (NIR-RGB), and Farmland (SAR-RGB). The experimental results demonstrate that DC-PanNet presents more accurate and robust performance than existing methods with less computational cost and fewer model parameters under extremely limited labeled sample conditions.