No AI summary available for this article.
Why It Matters
Adaptive optimization methods such as AdaGrad and Adam are widely used in modern neural-network training, but their adaptive scaling is primarily designed for vector-valued parameters and does not explicitly exploit matrix structure.
Provenance
Discovered via ArXiv and published by ArXiv.
Key Claims
Original description
Adaptive optimization methods such as AdaGrad and Adam are widely used in modern neural-network training, but their adaptive scaling is primarily designed for vector-valued parameters and does not explicitly exploit matrix structure. Recent matrix-aware optimizers demonstrate the benefits of structured optimization, yet a general theoretical framework for deriving matrix-aware adaptivity comparable to that of AdaGrad remains lacking. In this work, we develop a general Online Mirror Descent framework with adaptive proximal functions for matrix-valued parameters, providing a principled approach...
Discovered via ArXiv
Research papers and preprints from arXiv.
Publisher: arxiv.org
ID: http://arxiv.org/abs/2609.21815v1 · Indexed about 8 hours ago