Hide-and-Seek: A Template for Explainable AI

TL;DR

This paper introduces the Hide-and-Seek framework for training interpretable neural networks, providing a theoretical basis and demonstrating that high interpretability can be achieved without losing predictive accuracy.

Contribution

It presents a novel training framework for interpretable neural networks and offers a theoretical foundation for evaluating similar approaches.

Findings

Neural networks can be made highly interpretable without sacrificing accuracy

Theoretical analysis supports the effectiveness of the Hide-and-Seek framework

Experimental results validate the interpretability and performance of the proposed method

Abstract

Lack of transparency has been the Achilles heal of Neural Networks and their wider adoption in industry. Despite significant interest this shortcoming has not been adequately addressed. This study proposes a novel framework called Hide-and-Seek (HnS) for training Interpretable Neural Networks and establishes a theoretical foundation for exploring and comparing similar ideas. Extensive experimentation indicates that a high degree of interpretability can be imputed into Neural Networks, without sacrificing their predictive power.