Interpretable Disentanglement of Neural Networks by Extracting Class-Specific Subnetwork

2019/10/07 by Yulong Wang, Xiaolin Hu, Wang, Yulong +3
Computer Science · #Adversarial Robustness in Machine Learning #Anomaly Detection Techniques and Applications #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · pdf · doi:10.48550/arxiv.1910.02673

openalex publication_date 2019/10/07 · openalex created_date 2019/10/10 · openalex updated_date 2026/07/28

Abstract

We propose a novel perspective to understand deep neural networks in an interpretable disentanglement form. For each semantic class, we extract a class-specific functional subnetwork from the original full model, with compressed structure while maintaining comparable prediction performance. The structure representations of extracted subnetworks display a resemblance to their corresponding class semantic similarities. We also apply extracted subnetworks in visual explanation and adversarial example detection tasks by merely replacing the original full model with class-specific subnetworks. Experiments demonstrate that this intuitive operation can effectively improve explanation saliency accuracy for gradient-based explanation methods, and increase the detection rate for confidence score-based adversarial example detection methods.

Citations

• Distilling the Knowledge in a Neural Network
• Striving for Simplicity: The All Convolutional Net
• Axiomatic Attribution for Deep Networks
• Detecting Adversarial Samples from Artifacts
• SmoothGrad: removing noise by adding noise
• Methods for Interpreting and Understanding Deep Neural Networks
• Characterizing Adversarial Subspaces Using Local Intrinsic Dimensionality
• A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
• Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps

Related

• Visual Interpretability for Deep Learning: a Survey
• Unsupervised Learning of Neural Networks to Explain Neural Networks
• Architecture Disentanglement for Deep Neural Networks
• Unsupervised Learning of Neural Networks to Explain Neural Networks (extended abstract)
• Network Dissection: Quantifying Interpretability of Deep Visual\n Representations