iCaps: An Interpretable Classifier via Disentangled Capsule Networks

TL;DR

iCaps introduces an interpretable capsule network that disentangles class-relevant features and reduces overlap, providing transparent predictions with clear rationales without sacrificing accuracy.

Contribution

The paper presents a novel class-supervised disentanglement algorithm and regularizer to enhance interpretability of Capsule Networks.

Findings

iCaps achieves improved interpretability with clear rationales.

No performance degradation compared to existing Capsule Networks.

Effective on multiple datasets for image classification.

Abstract

We propose an interpretable Capsule Network, iCaps, for image classification. A capsule is a group of neurons nested inside each layer, and the one in the last layer is called a class capsule, which is a vector whose norm indicates a predicted probability for the class. Using the class capsule, existing Capsule Networks already provide some level of interpretability. However, there are two limitations which degrade its interpretability: 1) the class capsule also includes classification-irrelevant information, and 2) entities represented by the class capsule overlap. In this work, we address these two limitations using a novel class-supervised disentanglement algorithm and an additional regularizer, respectively. Through quantitative and qualitative evaluations on three datasets, we demonstrate that the resulting classifier, iCaps, provides a prediction along with clear rationales behind it with no performance degradation.