Explainable Deep Learning

Abstract

Recent advances in deep learning have made tremendous progress in the adoption of neural network models for tasks from resource utilization to autonomous driving. Most deep learning models are opaque black-box models that are not easily explainable. Unlike linear models, the weights of a neural network are not inherently interpretable to humans. The need for explainable deep learning has led to the development of a variety of methods that can help us better understand the decisions and decision-making process of neural network models. We note that many of the general post-hoc model-agnostic methods presented in Chap. 5 are applicable to deep learning models. This chapter presents a collection of explanation approaches that are specifically developed for neural networks by leveraging architecture or learning method.