Making Sense of CNNs: Interpreting Deep Representations & Their Invariances with INNs

TL;DR

This paper introduces an invertible method using INNs to interpret CNN representations by uncovering semantic concepts and invariances, enabling post-hoc understanding and modification of neural network models without performance loss.

Contribution

It presents a novel invertible approach that disentangles and visualizes learned invariances and semantic concepts in CNNs, enhancing interpretability.

Findings

Enables semantic understanding of CNN representations

Allows modification of learned invariances

Maintains model performance after interpretation

Abstract

To tackle increasingly complex tasks, it has become an essential ability of neural networks to learn abstract representations. These task-specific representations and, particularly, the invariances they capture turn neural networks into black box models that lack interpretability. To open such a black box, it is, therefore, crucial to uncover the different semantic concepts a model has learned as well as those that it has learned to be invariant to. We present an approach based on INNs that (i) recovers the task-specific, learned invariances by disentangling the remaining factor of variation in the data and that (ii) invertibly transforms these recovered invariances combined with the model representation into an equally expressive one with accessible semantic concepts. As a consequence, neural network representations become understandable by providing the means to (i) expose their semantic meaning, (ii) semantically modify a representation, and (iii) visualize individual learned semantic concepts and invariances. Our invertible approach significantly extends the abilities to understand black box models by enabling post-hoc interpretations of state-of-the-art networks without compromising their performance. Our implementation is available at https://compvis.github.io/invariances/ .