04:00
2026-09-21
machinebrief.com
machine-learning
Stiefel-AdamW: Geometry-Aware AdamW for Linear Factorization Blocks
Researchers introduced Stiefel-AdamW, a near drop-in replacement for the AdamW optimizer that constrains one factor of linear factorization blocks to the Stiefel manifold, according to arXiv paper 260…