Graph-Based Similarity of Neural Network Representations
This work addresses the challenge of interpreting black-box DNN representations for researchers in deep learning, though it appears incremental as it builds on existing similarity methods like CKA.
The authors tackled the problem of measuring similarity between neural network layer features by proposing Graph-Based Similarity (GBS), which constructs graphs from hidden layer outputs and shows state-of-the-art performance in reflecting similarity and providing insights into adversarial sample behavior.
Understanding the black-box representations in Deep Neural Networks (DNN) is an essential problem in deep learning. In this work, we propose Graph-Based Similarity (GBS) to measure the similarity of layer features. Contrary to previous works that compute the similarity directly on the feature maps, GBS measures the correlation based on the graph constructed with hidden layer outputs. By treating each input sample as a node and the corresponding layer output similarity as edges, we construct the graph of DNN representations for each layer. The similarity between graphs of layers identifies the correspondences between representations of models trained in different datasets and initializations. We demonstrate and prove the invariance property of GBS, including invariance to orthogonal transformation and invariance to isotropic scaling, and compare GBS with CKA. GBS shows state-of-the-art performance in reflecting the similarity and provides insights on explaining the adversarial sample behavior on the hidden layer space.