# The Geometry of Spectral Gradient Descent: Layerwise Criteria for SignSGD vs SpecSGD

Canonical HTML: https://hiroki11x.github.io/ja/publications/geometry-of-spectral-gradient-descent/

- 著者: Hiroki Naganuma, Laura Gomezjurado Gonzalez, Mahdi Ghaznavi, Ioannis Mitliagkas
- 掲載先: ICLR 2026 Workshop on Geometry-grounded Representation Learning and Generative Modeling (ICLR 2026 Workshop)
- 日付: 2026-02
- 研究領域: [機械学習の最適化](https://hiroki11x.github.io/ja/research/#optimization)

## 概要

概要は原文リンクをご覧ください。

## 一次資料

- [OpenReview](https://openreview.net/forum?id=lDs4vkSX7J)