Towards Understanding Learning Representations: To What Extent Do Different Neural Networks Learn the Same Representation

2018/10/27 by Liwei Wang, Wang, Liwei, Lunjia Hu +12 · 14 citations
Computer Science · Mathematics · #Domain Adaptation and Few-Shot Learning #Machine Learning and Data Classification #Neural Networks and Applications #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.1810.11750

17 pages, 6 figures

arxiv created 2018/11/28 · arxiv updated 2018/11/30

Abstract

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.

Citations

• Very Deep Convolutional Networks for Large-Scale Image Recognition

Cited by

• Towards Understanding Layer Contributions in Tabular In-Context Learning Models
• To Align or Not to Align: Strategic Multimodal Representation Alignment for Optimal Performance
• Understanding the Dynamics of DNNs Using Graph Modularity
• Representation Potentials of Foundation Models for Multimodal Alignment: A Survey
• Emerging Cross-lingual Structure in Pretrained Language Models
• Representation Transfer by Optimal Transport
• Non-uniqueness phenomenon of object representation in modelling IT cortex by deep convolutional neural network (DCNN)
• On Geometric Structure of Activation Spaces in Neural Networks
• ImageNet Pre-training also Transfers Non-Robustness
• When Does Closeness in Distribution Imply Representational Similarity? An Identifiability Perspective
• Linear Representation Transferability Hypothesis: Leveraging Small Models to Steer Large Models
• On the geometry of generalization and memorization in deep neural networks
• Confidence Calibration for Convolutional Neural Networks Using Structured Dropout
• Neuron Populations Exhibit Divergent Selectivity with Scale

Related

• Subspace Match Probably Does Not Accurately Assess the Similarity of Learned Representations
• When Representations Align: Universality in Representation Learning Dynamics
• Convergent Learning: Do different neural networks learn the same representations?
• Similarity and Matching of Neural Network Representations
• How Do the Architecture and Optimizer Affect Representation Learning? On the Training Dynamics of Representations in Deep Neural Networks