A Framework to Learn with Interpretation

TL;DR

This paper introduces a new framework for jointly learning predictive and interpretability models in deep learning, balancing accuracy with human-understandable explanations through specialized architecture and regularization.

Contribution

It proposes a novel architecture and regularization approach to produce interpretable models with minimal accuracy loss, applicable both during training and post-hoc to pre-trained networks.

Findings

Effective in providing local and global interpretability

Maintains high predictive accuracy while enhancing interpretability

Outperforms several state-of-the-art interpretability methods

Abstract

To tackle interpretability in deep learning, we present a novel framework to jointly learn a predictive model and its associated interpretation model. The interpreter provides both local and global interpretability about the predictive model in terms of human-understandable high level attribute functions, with minimal loss of accuracy. This is achieved by a dedicated architecture and well chosen regularization penalties. We seek for a small-size dictionary of high level attribute functions that take as inputs the outputs of selected hidden layers and whose outputs feed a linear classifier. We impose strong conciseness on the activation of attributes with an entropy-based criterion while enforcing fidelity to both inputs and outputs of the predictive model. A detailed pipeline to visualize the learnt features is also developed. Moreover, besides generating interpretable models by design, our approach can be specialized to provide post-hoc interpretations for a pre-trained neural network. We validate our approach against several state-of-the-art methods on multiple datasets and show its efficacy on both kinds of tasks.