Towards Frequency-Based Explanation for Robust CNN

Abstract

Current explanation techniques towards a transparent Convolutional Neural Network (CNN) mainly focuses on building connections between the human-understandable input features with models' prediction, overlooking an alternative representation of the input, the frequency components decomposition. In this work, we present an analysis of the connection between the distribution of frequency components in the input dataset and the reasoning process the model learns from the data. We further provide quantification analysis about the contribution of different frequency components toward the model's prediction. We show that the vulnerability of the model against tiny distortions is a result of the model is relying on the high-frequency features, the target features of the adversarial (black and white-box) attackers, to make the prediction. We further show that if the model develops stronger association between the low-frequency component with true labels, the model is more robust, which is the explanation of why adversarially trained models are more robust against tiny distortions.

References in corpus (11)

• Explaining and Harnessing Adversarial Examples
• Intriguing properties of neural networks
• Deep Residual Learning for Image Recognition
• Axiomatic Attribution for Deep Networks
• Learning Important Features Through Propagating Activation Differences
• Understanding Black-box Predictions via Influence Functions
• Visualizing and Understanding Recurrent Networks
• Certified Adversarial Robustness via Randomized Smoothing
• Visualizing and Understanding Convolutional Networks
• Representer Point Selection for Explaining Deep Neural Networks
• High Frequency Component Helps Explain the Generalization of Convolutional Neural Networks

Cited by in corpus (5)

• Robust Data Hiding Using Inverse Gradient Attention
• Learning Frequency Domain Approximation for Binary Neural Networks
• Human-interpretable model explainability on high-dimensional data
• A Singular Value Perspective on Model Robustness
• Fourier Transform Approximation as an Auxiliary Task for Image Classification