CHIP: Channel-wise Disentangled Interpretation of Deep Convolutional Neural Networks

Xinrui Cui , Dan Wang, Z. Jane Wang
University of British Columbia
PDF
Model visualization

The workflow of the proposed CHIP model. In the first stage, we learn the CHIP model based on the perturbed dataset which contains around a hundred million perturbed networks for the overall 1000 classes in ILSVRC 2015 dataset. After the optimization, we obtain the distilled class-discriminative importance of channels for different classes. In the second stage, given a specific image and the class of interest (banana and lemon), the class-discriminative visual interpretation is obtained for the target class by utilizing the distilled knowledge.

Abstract

With the increasing popularity of deep convolutional neural networks (DCNNs), in addition to achieving high accuracy, it becomes increasingly important to explain how DCNNs make their decisions. In this work, we propose a CHannel-wise disentangled InterPretation (CHIP) model for visual interpretations of DCNN predictions. The proposed model distills the class-discriminative importance of channels in DCNN by utilizing sparse regularization. We first introduce network perturbation to learn the CHIP model. The proposed model is capable to not only distill the global perspective knowledge from networks but also present class-discriminative visual interpretations for the predictions of networks. It is noteworthy that the CHIP model is able to interpret different layers of networks without re-training. By combining the distilled interpretation knowledge at different layers, we further propose the Refined CHIP visual interpretation that is both high-resolution and class-discriminative. Based on qualitative and quantitative experiments on different datasets and networks, the proposed model provides promising visual interpretations for network predictions in image classification task compared with existing visual interpretation methods. The proposed model also outperforms related approaches in the ILSVRC 2015 weakly-supervised localization task.

BibTeX

@ARTICLE{8924894,
  author={Cui, Xinrui and Wang, Dan and Wang, Z. Jane},
  journal={IEEE Transactions on Neural Networks and Learning Systems}, 
  title={CHIP: Channel-Wise Disentangled Interpretation of Deep Convolutional Neural Networks}, 
  year={2020},
  volume={31},
  number={10},
  pages={4143-4156},
  doi={10.1109/TNNLS.2019.2952322}
}