CS seminar

Title: Designing for reliability: Fairness and interpretability aware vision transformers

Speaker

Dr. Yao Qiang, Assistant Professor, Computer Science, Oakland University

Abstract

Vision Transformer (ViT) has recently gained significant attention in solving computer vision (CV) problems due to its capability of extracting informative features and modeling long-range dependencies through the attention mechanism. Whereas recent works have explored the trustworthiness of ViT, the fairness and interpretability of ViTs have not kept pace with their promising performance. To improve the fairness of ViTs, we design a fairness-aware algorithm and develop a new framework via Debiased Self-Attention (DSA). DSA enforces ViT to eliminate spurious features correlated with the sensitive label for bias mitigation and simultaneously retain real features for target prediction. This leads to improved fairness guarantees over prior works on multiple prediction tasks without compromising target prediction performance. Instead of developing another post-hoc explanation approach, we introduce a novel training procedure that inherently enhances ViT's interpretability. Our interpretability-aware ViT (IA-ViT) comprises a feature extractor, a predictor, and an interpreter, which are trained jointly with an interpretability-aware training objective. Consequently, the interpreter simulates the behavior of the predictor and provides a faithful explanation through its single-head self-attention mechanism. Our comprehensive experimental results demonstrate the effectiveness of IA-ViT in several image classification tasks.

Bio

Yao Qiang joins Oakland University as an assistant professor in the Computer Science and Engineering department this year.  He graduated from the Department of Computer Science at Wayne State University, working in the Trustworthy AI lab under the supervision of Dr. Dongxiao Zhu. His research mainly focuses on Trustworthy AI, Large Language Models, and Machine Learning Theory and Application. Yao's dedication to these areas has culminated in the publication of numerous research papers at the most competitive AI conferences, including NeurIPS, ICML, EACL, ECCV, AAAI, IJCAI, etc. His passion for research drives him to delve deeper into the frontiers of science and encourages him to transform theoretical discoveries into practical innovations that make a meaningful impact on society.