No confident code implementation found for this paper.
We compute the minimum-norm weights of a deep linear ResNet, and find that the inductive bias of this architecture lies between minimizing nuclear norm and rank. This implies that, with appropriate hyperparameters, deep nonlinear ResNets have an inductive bias towards minimizing bottleneck rank.
No confident code match yet
We couldn't find an author-owned or strongly-evidenced community implementation for this paper. Any repos shown below are weak matches — verify before relying on them.
No code implementations found yet.
Know of an implementation? Let us know in the comments below!
Theoretical Analysis of Inductive Biases in Deep Convolutional Networks
Bottleneck Structure in Learned Features: Low-Dimension vs Regularity Tradeoff
Provable Emergence of Deep Neural Collapse and Low-Rank Bias in L²-Regularized Nonlinear Networks
Low-Rank Learning by Design: the Role of Network Architecture and Activation Linearity in Gradient Rank Collapse
Deep Learning as a Convex Paradigm of Computation: Minimizing Circuit Size with ResNets
Generalization Bounds for Rank-sparse Neural Networks
@article{boixadsera2025inductive,
title = {On the inductive bias of infinite-depth ResNets and the bottleneck rank},
author = {Enric Boix-Adsera},
year = {2025},
eprint = {2501.19149},
archivePrefix = {arXiv},
url = {http://arxiv.org/abs/2501.19149v1}
}Shows whether official code is linked for this paper.
[](https://codeofpaper.com/paper/2501.19149)