Detecting Deep Neural Network Defects with Data Flow Analysis
This addresses debugging challenges for DNN developers, but appears incremental as it builds on existing data flow analysis techniques.
The paper tackles the problem of distinguishing between inevitable low precision and defects in deep neural networks (DNNs) by using internal data flow analysis to locate root causes, resulting in a tool called DeepMorph that guides developers to improve models.
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.