Wenda Wei

Wenda Wei

Auto-generated profile, based on publicly available information

Wenda Wei

National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, Jiangsu, China

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

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

Representative Publications

View All Publications
18 August 2023 in IT Professional
Verified
Journal Article
Closed access
WWWenda Wei
Wenda Wei
CWChenyang Wang
Chenyang Wang
XZXinyang Zheng
Xinyang Zheng
HYHengshan Yue
Hengshan Yue
Today, various domain-specific convolutional neural network (CNN) accelerators are deployed in large-scale systems to satisfy the massive computational demands of current deep CNNs. Although bringing significant performance improvements, the highly integrated CNN accelerators are more susceptible to faults caused by radiation, aging, and process variation. CNNs have been increasingly deployed in security-critical areas, requiring more attention to reliable execution. Although the classical fault-tolerant approaches are error-effective, the performance/energy overheads introduced are nonnegligible, which is the opposite of CNN accelerator design philosophy. In this article, we leverage CNN’s intrinsic tolerance for minor errors to explore approximate fault-tolerance (ApFT) opportunities for CNN accelerator fault-tolerance overhead reduction. Specifically, we discuss two branches of ApFT designs: selective duplicating-based approximate fault tolerance (S-ApFT) and imprecise checking-based approximate fault tolerance (I-ApFT). The results show that S-ApFT and I-ApFT can achieve comparable error-detection ability and dual-modular redundancy while achieving significant performance improvements.

17 June 2021 in The Journal of the Textile Institute
Verified
Journal Article
Closed access
XZXiaoping Zheng
Xiaoping Zheng
CLChengxia Liu
Chengxia Liu
WWWenda Wei
Wenda Wei
Garment bagging is the plastic deformation of the knee, elbow and other parts of the garment when the human body moves. The current methods and indexes for evaluating the bagging performance of woven garments are not comprehensive enough. In view of this situation, 16 pieces of fabrics were made into pants-tube for actual wearing and bagging. The camera and the 3D laser scanner were used to acquire images and point cloud data of garments respectively. Then feature extraction was performed on the processed data to get two traditional geometric quantization indexes, bagging height and bagging volume. Furthermore, two new geometric quantification indexes were proposed: warp and weft maximum bagging rate. And several kinds of surface quantization indexes were also could be extracted, namely mean, average deviation, roughness, variance and torsion. At the same time, the subjective evaluation results had been obtained by the bagging images and the bagging 3D curved surfaces. After that, correlation analysis and ridge regression analysis were carried out between the two groups of indexes and the subjective score results. The results showed that, the two new geometric quantitative indexes were highly correlated with the subjective score, and the correlation coefficient were −0.883 and −0.923, respectively. The surface quantization indexes regression equation and the geometric quantification index regression equation were both accurate in predicting the subjective score, and the adjusted R2 reached 92.7% and 91.3%. The 3D laser scanning technology could be used to evaluate the garment bagging accurately.

8 August 2026
Verified
Conference Paper
Open access
Yufeng LiYufeng Li
Yufeng Li
XYXiao-Wen Yang
Xiao-Wen Yang
WWWen-Da Wei
Wen-Da Wei
JSJie-Jing Shao
Jie-Jing Shao
LGLan-Zhe Guo
Lan-Zhe Guo
Recent advancements in neuro-symbolic learning (NeSy) have shown significant promise in integrating deep learning with symbolic reasoning, offering both interpretability and generalization. However, the prevalence of reasoning shortcuts, where the NeSy system predicts incorrect intermediate concepts while maintaining high final accuracy, poses a substantial challenge. This is especially problematic in domains requiring reliable and transparent decision-making. Inspired by recent theories, we find that existing methods fail to address the reasoning shortcut issue when the knowledge base lacks sufficient complexity, highlighting their vulnerability in real-world applications. In this work, we present a novel method called DKA to address this issue. It introduces a limited set of concept-supervised data to enhance the knowledge base, effectively solving the reasoning shortcut problem and improving the applicability of the NeSy system. Theoretical analysis reveals that DKA can reduce shortcut risks with improved data efficiency. Empirical studies across multiple tasks within various neuro-symbolic frameworks also verify the effectiveness of the DKA method.