Towards Understanding Learning Representations: To What Extent Do Different Neural Networks Learn the Same Representation
This work addresses a foundational gap in understanding neural network representations for the machine learning community, though it is incremental as it moves a tiny step towards a broader theory.
The paper tackles the problem of quantifying how similar the learned representations are in neural networks with identical architectures but different initializations, finding that convolutional layers are less similar than expected in terms of subspace match.
It is widely believed that learning good representations is one of the main reasons for the success of deep neural networks. Although highly intuitive, there is a lack of theory and systematic approach quantitatively characterizing what representations do deep neural networks learn. In this work, we move a tiny step towards a theory and better understanding of the representations. Specifically, we study a simpler problem: How similar are the representations learned by two networks with identical architecture but trained from different initializations. We develop a rigorous theory based on the neuron activation subspace match model. The theory gives a complete characterization of the structure of neuron activation subspace matches, where the core concepts are maximum match and simple match which describe the overall and the finest similarity between sets of neurons in two networks respectively. We also propose efficient algorithms to find the maximum match and simple matches. Finally, we conduct extensive experiments using our algorithms. Experimental results suggest that, surprisingly, representations learned by the same convolutional layers of networks trained from different initializations are not as similar as prevalently expected, at least in terms of subspace match.