Residual Alignment: Uncovering the Mechanisms of Residual Networks

TL;DR

This paper empirically investigates the mechanisms behind ResNet's success, revealing a phenomenon called Residual Alignment that correlates with good generalization and involves geometric and spectral properties of residual Jacobians.

Contribution

It introduces the concept of Residual Alignment, characterizes its properties, and demonstrates its occurrence across various architectures and datasets, providing insights into ResNet's effectiveness.

Findings

Residual Alignment occurs in well-generalizing models

Residual Jacobians are low-rank and their singular vectors align

Residual Jacobian singular values scale inversely with depth

Abstract

The ResNet architecture has been widely adopted in deep learning due to its significant boost to performance through the use of simple skip connections, yet the underlying mechanisms leading to its success remain largely unknown. In this paper, we conduct a thorough empirical study of the ResNet architecture in classification tasks by linearizing its constituent residual blocks using Residual Jacobians and measuring their singular value decompositions. Our measurements reveal a process called Residual Alignment (RA) characterized by four properties: (RA1) intermediate representations of a given input are equispaced on a line, embedded in high dimensional space, as observed by Gai and Zhang [2021]; (RA2) top left and right singular vectors of Residual Jacobians align with each other and across different depths; (RA3) Residual Jacobians are at most rank C for fully-connected ResNets, where C is the number of classes; and (RA4) top singular values of Residual Jacobians scale inversely with depth. RA consistently occurs in models that generalize well, in both fully-connected and convolutional architectures, across various depths and widths, for varying numbers of classes, on all tested benchmark datasets, but ceases to occur once the skip connections are removed. It also provably occurs in a novel mathematical model we propose. This phenomenon reveals a strong alignment between residual branches of a ResNet (RA2+4), imparting a highly rigid geometric structure to the intermediate representations as they progress linearly through the network (RA1) up to the final layer, where they undergo Neural Collapse.