Towards glass-box CNNs

TL;DR

This paper proposes a transparent, binary prototype of CNNs that enhances interpretability and class separability by unfolding and repacking conventional CNNs into a three-layer model with invariant and equivariant representations.

Contribution

It introduces a novel, analytically transparent CNN prototype that improves interpretability and feature separability for large-scale applications.

Findings

Repacked CNN with a three-layer model improves interpretability.

Representation layer captures class invariance and symmetry transformations.

Comparison shows improved feature separability over AlexNet.

Abstract

With the substantial performance of neural networks in sensitive fields increases the need for interpretable deep learning models. Major challenge is to uncover the multiscale and distributed representation hidden inside the basket mappings of the deep neural networks. Researchers have been trying to comprehend it through visual analysis of features, mathematical structures, or other data-driven approaches. Here, we work on implementation invariances of CNN-based representations and present an analytical binary prototype that provides useful insights for large scale real-life applications. We begin by unfolding conventional CNN and then repack it with a more transparent representation. Inspired by the attainment of neural networks, we choose to present our findings as a three-layer model. First is a representation layer that encompasses both the class information (group invariant) and symmetric transformations (group equivariant) of input images. Through these transformations, we decrease intra-class distance and increase the inter-class distance. It is then passed through a dimension reduction layer followed by a classifier. The proposed representation is compared with the equivariance of AlexNet (CNN) internal representation for better dissemination of simulation results. We foresee following immediate advantages of this toy version: i) contributes pre-processing of data to increase the feature or class separability in large scale problems, ii) helps designing neural architecture to improve the classification performance in multi-class problems, and iii) helps building interpretable CNN through scalable functional blocks.