-
- UCHIBE Eiji
- ATR Computational Neuroscience Labs.
Bibliographic Information
- Other Title
-
- 方策探査法のための多重重点サンプリングを用いた経験再利用
Abstract
<p>In policy search methods, importance sampling is widely used to reutilize samples drawn from previous sampling distributions that are usually different from the current one. Previous studies create uniform mixtures of previous sampling distributions as proposal distribution. To further improve sample efficiency, we introduce adaptive multiple importance sampling that optimizes the mixing coefficients to minimize the variance of the importance sampling estimator. We apply the proposed method to several policy search methods and experimental results on some benchmark control tasks show that all the methods improve sample efficiency.</p>
Journal
-
- The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec)
-
The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) 2018 (0), 1A1-E13-, 2018
The Japan Society of Mechanical Engineers
- Tweet
Details 詳細情報について
-
- CRID
- 1390282763080204672
-
- NII Article ID
- 130007550888
-
- ISSN
- 24243124
-
- Text Lang
- ja
-
- Data Source
-
- JaLC
- Crossref
- CiNii Articles
-
- Abstract License Flag
- Disallowed