Effects of Data Geometry in Early Deep LearningDownload PDF

Published: 28 Jan 2022, Last Modified: 13 Feb 2023ICLR 2022 SubmittedReaders: Everyone
Keywords: Deep learning, geometry, manifolds, deep learning theory
Abstract: Deep neural networks can approximate functions on different types of data, from images to graphs, with varied underlying structure.This underlying structure can be viewed as the geometry of the data manifold. By extending recent advances in the theoretical understanding of neural networks, we study how a randomly initialized neural network with piecewise linear activation splits the data manifold into regions where the neural network behaves as a linear function. We derive bounds on the number of linear regions and the distance to boundaries of these linear regions on the data manifold. This leads to insights into the expressivity of randomly initialized deep neural networks on non-Euclidean data sets. We empirically corroborate our theoretical results using a toy supervised learning problem. Our experiments demonstrate that number of linear regions varies across manifolds and how our results hold upon changing neural network architectures. We further demonstrate how the complexity of linear regions changes on the low dimensional manifold of images as training progresses, using the MetFaces dataset.
One-sentence Summary: We provide theoretical results for understanding the capacity of neural networks in light of the manifold hypothesis and corroborate them with experiments.
Supplementary Material: zip
9 Replies

Loading