Scheduling of Damping in Natural Gradient Method
-
- NAGANUMA Hiroki
- University of Montreal Mila
-
- FUJIMORI Gaku
- Tokyo University of Science
-
- TAKEUCHI Mari
- University College London
-
- NAGASE Jumpei
- Shibaura Institute of Technology
Bibliographic Information
- Other Title
-
- 自然勾配学習法における Damping 項のスケジューリング
Abstract
<p>In recent years, second-order optimization with a fast convergence rate has been used in deep learning owing to fast approximation methods for natural gradient methods. Second-order optimization requires the inverse computation of the information matrix, which generally degenerates in the deep learning problem. Therefore, as a heuristic, a damping method adds a unit matrix multiplied by a constant. This study proposed a method for scheduling damping motivated by the Levenberg-Marquardt method for determining damping and investigated its effectiveness.</p>
Journal
-
- Proceedings of the Annual Conference of JSAI
-
Proceedings of the Annual Conference of JSAI JSAI2022 (0), 1D1GS202-1D1GS202, 2022
The Japanese Society for Artificial Intelligence
- Tweet
Details 詳細情報について
-
- CRID
- 1390292706081864320
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
-
- Abstract License Flag
- Disallowed