BatchNormalizationProvablyAvoidsRankCollapse forRandomlyInitialisedDeepNetworks

Feb-10-2026, 14:14:55 GMT–Neural Information Processing Systems

Given the rich literature on random matrices, it is not surprising to find that the rank of the intermediate representations in unnormalized networks collapses quickly with depth.

artificial intelligence, arxivpreprintarxiv, machine learning, (16 more...)

Neural Information Processing Systems

Feb-10-2026, 14:14:55 GMT

Conferences PDF

Add feedback

Country:
- North America > Canada > British Columbia > Metro Vancouver Regional District > Vancouver (0.04)

Technology:
- Information Technology > Artificial Intelligence > Machine Learning > Neural Networks (1.00)

Duplicate Docs Excel Report

Title
Batch Normalization Provably Avoids Rank Collapse for Randomly Initialised Deep Networks

Similar Docs Excel Report more

Title	Similarity	Source
None found