This paper proposes a method to visualize the discrimination power of intermediate-layer visual patterns encoded by a DNN. Specifically, we visualize (1) how the DNN gradually learns regional visual patterns in each intermediate layer during the training process, and (2) the effects of the DNN using non-discriminative patterns in low layers to construct disciminative patterns in middle/high layers through the forward propagation. Based on our visualization method, we can quantify knowledge points (i.e. the number of discriminative visual patterns) learned by the DNN to evaluate the representation capacity of the DNN. Furthermore, this method also provides new insights into signal-processing behaviors of existing deep-learning techniques, such as adversarial attacks and knowledge distillation.
Mingjie Li (Shanghai Jiao Tong University)
Shaobo Wang (Harbin Institute of Technology)
Shaobo Wang is now a 4th year undergraduate student major in Software Engineering at Harbin Institute of Technology. His research interests include eXplainable AI(XAI), Self-Supervised Learning and Cross-Modal Learning.
Quanshi Zhang (University of Tokyo)
More from the Same Authors
2021 Poster: Interpreting Representation Quality of DNNs for 3D Point Cloud Processing »
Wen Shen · Qihan Ren · Dongrui Liu · Quanshi Zhang
2021 Poster: Towards a Unified Game-Theoretic View of Adversarial Perturbations and Robustness »
Jie Ren · Die Zhang · Yisen Wang · Lu Chen · Zhanpeng Zhou · Yiting Chen · Xu Cheng · Xin Wang · Meng Zhou · Jie Shi · Quanshi Zhang