Causal Learning and Explanation of Deep Neural Networks via Autoencoded Activations

TL;DR

This paper introduces a method to explain deep neural network predictions by extracting human-understandable concepts with autoencoders, then building a causal model to identify influential features.

Contribution

It presents a novel approach combining autoencoders and causal modeling to interpret CNN decisions in a human-understandable way.

Findings

Successfully extracts salient concepts from CNN activations.

Builds a Bayesian causal model for explanation.

Visualizes features with causal influence on classification.

Abstract

Deep neural networks are complex and opaque. As they enter application in a variety of important and safety critical domains, users seek methods to explain their output predictions. We develop an approach to explaining deep neural networks by constructing causal models on salient concepts contained in a CNN. We develop methods to extract salient concepts throughout a target network by using autoencoders trained to extract human-understandable representations of network activations. We then build a bayesian causal model using these extracted concepts as variables in order to explain image classification. Finally, we use this causal model to identify and visualize features with significant causal influence on final classification.