Finding Representative Interpretations on Convolutional Neural Networks

TL;DR

This paper introduces an unsupervised method to generate highly representative interpretations of CNN decision logic across large groups of similar images, enhancing understanding and generalization.

Contribution

It formulates the interpretation task as a co-clustering problem and develops a novel approach to produce representative explanations for CNNs on multiple images.

Findings

Effective interpretation of CNNs on large image groups

Superior performance demonstrated through extensive experiments

Provides visualization and similarity ranking methods

Abstract

Interpreting the decision logic behind effective deep convolutional neural networks (CNN) on images complements the success of deep learning models. However, the existing methods can only interpret some specific decision logic on individual or a small number of images. To facilitate human understandability and generalization ability, it is important to develop representative interpretations that interpret common decision logics of a CNN on a large group of similar images, which reveal the common semantics data contributes to many closely related predictions. In this paper, we develop a novel unsupervised approach to produce a highly representative interpretation for a large number of similar images. We formulate the problem of finding representative interpretations as a co-clustering problem, and convert it into a submodular cost submodular cover problem based on a sample of the linear decision boundaries of a CNN. We also present a visualization and similarity ranking method. Our extensive experiments demonstrate the excellent performance of our method.