vix.ing · top · new · best · stats · spec

Bottleneck Structure in Learned Features: Low-Dimension vs Regularity Tradeoff

2023/05/30 by Arthur Paul Jacot, Jacot, Arthur · 1 citation
Computer Science · Engineering · Physics and Astronomy · #Domain Adaptation and Few-Shot Learning #Sparse and Compressive Sensing Techniques #Model Reduction and Neural Networks

paper · pdf · doi:10.48550/arxiv.2305.19008

Abstract

Previous work has shown that DNNs with large depth L and L2-regularization are biased towards learning low-dimensional representations of the inputs, which can be interpreted as minimizing a notion of rank R(0)(f) of the learned function f, conjectured to be the Bottleneck rank. We compute finite depth corrections to this result, revealing a measure R(1) of regularity which bounds the pseudo-determinant of the Jacobian |Jf(x)|+ and is subadditive under composition and addition. This formalizes a balance between learning low-dimensional representations and minimizing complexity/irregularity in the feature maps, allowing the network to learn the `right' inner dimension. Finally, we prove the conjectured bottleneck structure in the learned features as L→∞: for large depths, almost all hidden representations are approximately R(0)(f)-dimensional, and almost all weight matrices W have R(0)(f) singular values close to 1 while the others are O(L-(1)/(2)). Interestingly, the use of large learning rates is required to guarantee an order O(L) NTK which in turns guarantees infinite depth convergence of the representations of almost all layers.

Cited by

Related