Timezone: »

A Theoretical Explanation for Perplexing Behaviors of Backpropagation-based Visualizations
Weili Nie · Yang Zhang · Ankit Patel

Fri Jul 13 08:40 AM -- 08:50 AM (PDT) @ K1

Backpropagation-based visualizations have been proposed to interpret convolutional neural networks (CNNs), however a theory is missing to justify their behaviors: Guided backpropagation (GBP) and deconvolutional network (DeconvNet) generate more human-interpretable but less class-sensitive visualizations than saliency map. Motivated by this, we develop a theoretical explanation revealing that GBP and DeconvNet are essentially doing (partial) image recovery which is unrelated to the network decisions. Specifically, our analysis shows that the backward ReLU introduced by GBP and DeconvNet, and the local connections in CNNs are the two main causes of compelling visualizations. Extensive experiments are provided that support the theoretical analysis.

Author Information

Weili Nie (Rice University)
Yang Zhang (Rice University)
Ankit Patel (Rice University, Baylor College of Medicine)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors