Synthesizing the preferred inputs for neurons in neural networks via deep generator networks

TL;DR

This paper enhances activation maximization for neural network interpretability by using deep generator networks, producing realistic images that reveal neuron features and generalize across datasets and architectures.

Contribution

Introduces a novel method combining deep generator networks with activation maximization to produce high-quality, interpretable neuron visualizations that generalize well.

Findings

Generated images are almost real and highly interpretable.

Method generalizes to new datasets and architectures.

Produces creative and recognizable images.

Abstract

Deep neural networks (DNNs) have demonstrated state-of-the-art results on many pattern recognition tasks, especially vision classification problems. Understanding the inner workings of such computational brains is both fascinating basic science that is interesting in its own right - similar to why we study the human brain - and will enable researchers to further improve DNNs. One path to understanding how a neural network functions internally is to study what each of its neurons has learned to detect. One such method is called activation maximization (AM), which synthesizes an input (e.g. an image) that highly activates a neuron. Here we dramatically improve the qualitative state of the art of activation maximization by harnessing a powerful, learned prior: a deep generator network (DGN). The algorithm (1) generates qualitatively state-of-the-art synthetic images that look almost real, (2) reveals the features learned by each neuron in an interpretable way, (3) generalizes well to new datasets and somewhat well to different network architectures without requiring the prior to be relearned, and (4) can be considered as a high-quality generative method (in this case, by generating novel, creative, interesting, recognizable images).