Detecting Deep Neural Network Defects with Data Flow Analysis

TL;DR

This paper introduces DeepMorph, a data flow analysis tool that helps identify whether low precision in deep neural networks is due to inherent limitations or defects, aiding developers in model improvement.

Contribution

The paper presents a novel data flow analysis approach, DeepMorph, for diagnosing the root causes of low precision in DNNs, which was not addressed in prior work.

Findings

DeepMorph effectively identifies defect causes in DNNs.

Data flow footprints reveal insights into model precision issues.

Guides developers to improve neural network models.

Abstract

Deep neural networks (DNNs) are shown to be promising solutions in many challenging artificial intelligence tasks. However, it is very hard to figure out whether the low precision of a DNN model is an inevitable result, or caused by defects. This paper aims at addressing this challenging problem. We find that the internal data flow footprints of a DNN model can provide insights to locate the root cause effectively. We develop DeepMorph (DNN Tomography) to analyze the root cause, which can guide a DNN developer to improve the model.