Shared Representational Geometry Across Neural Networks

Abstract

Different neural networks trained on the same dataset often learn similar input-output mappings with very different weights. Is there some correspondence between these neural network solutions? For linear networks, it has been shown that different instances of the same network architecture encode the same representational similarity matrix, and their neural activity patterns are connected by orthogonal transformations. However, it is unclear if this holds for non-linear networks. Using a shared response model, we show that different neural networks encode the same input examples as different orthogonal transformations of an underlying shared representation. We test this claim using both standard convolutional neural networks and residual networks on CIFAR10 and CIFAR100.

Integration of Deep Learning Theories workshop, NeurIPS 2018

References in corpus (5)

• Identifying and attacking the saddle point problem in high-dimensional non-convex optimization
• SVCCA: Singular Vector Canonical Correlation Analysis for Deep Learning Dynamics and Interpretability
• Convergent Learning: Do different neural networks learn the same representations?
• Enabling Factor Analysis on Thousand-Subject Neuroimaging Datasets
• $\mathcal{G}$-SGD: Optimizing ReLU Neural Networks in its Positively Scale-Invariant Space

Cited by in corpus (1)

• High-contrast "gaudy" images improve the training of deep neural network models of visual cortex