Understanding Neural Networks Through Deep Visualization

TLDR

Recent advances in training deep neural networks, especially convolutional nets for image recognition, have outpaced our understanding of their internal computations, motivating the need for improved visualization tools. The authors introduce two new visualization tools for deep neural networks. They provide two open‑source tools: one that displays per‑layer activations for images or live video, and another that visualizes layer features via regularized image‑space optimization with enhanced regularization for clearer, more interpretable visualizations. Live activation visualizations help users build intuition about how convolutional nets process inputs.

Abstract

Recent years have produced great advances in training large, deep neural networks (DNNs), including notable successes in training convolutional neural networks (convnets) to recognize natural images. However, our understanding of how these models work, especially what computations they perform at intermediate layers, has lagged behind. Progress in the field will be further accelerated by the development of better tools for visualizing and interpreting neural nets. We introduce two such tools here. The first is a tool that visualizes the activations produced on each layer of a trained convnet as it processes an image or video (e.g. a live webcam stream). We have found that looking at live activations that change in response to user input helps build valuable intuitions about how convnets work. The second tool enables visualizing features at each layer of a DNN via regularized optimization in image space. Because previous versions of this idea produced less recognizable images, here we introduce several new regularization methods that combine to produce qualitatively clearer, more interpretable visualizations. Both tools are open source and work on a pre-trained convnet with minimal setup.