Towards Efficient Visual Simplification of Computational Graphs in Deep Neural Networks

TL;DR

This paper introduces a set of visualization techniques and an interactive system to simplify and analyze large-scale computational graphs in deep neural networks, improving clarity and diagnostic efficiency.

Contribution

It presents novel visual simplification methods and an interactive visualization system integrated into an open-source toolkit for large-scale DNN graphs.

Findings

Reduces graph elements by 60% on average

Enhances model recognition and diagnosis efficiency

Supports graphs with up to 10,000 elements

Abstract

A computational graph in a deep neural network (DNN) denotes a specific data flow diagram (DFD) composed of many tensors and operators. Existing toolkits for visualizing computational graphs are not applicable when the structure is highly complicated and large-scale (e.g., BERT [1]). To address this problem, we propose leveraging a suite of visual simplification techniques, including a cycle-removing method, a module-based edge-pruning algorithm, and an isomorphic subgraph stacking strategy. We design and implement an interactive visualization system that is suitable for computational graphs with up to 10 thousand elements. Experimental results and usage scenarios demonstrate that our tool reduces 60% elements on average and hence enhances the performance for recognizing and diagnosing DNN models. Our contributions are integrated into an open-source DNN visualization toolkit, namely, MindInsight [2].