Paper Abstract and Keywords |
Presentation |
2018-06-13 10:00
Active Level Set Estimation with Multi-fidelity Evaluations Shion Takeno (Nitech), Hitoshi Fukuoka (Nagoya Univ.), Yuhki Tsukada (Nagoya Univ./JST), Toshiyuki Koyama (Nagoya Univ.), Motoki Shiga (Gifu Univ./JST/RIKEN), Ichiro Takeuchi (NITech/NIMS/RIKEN), Masayuki Karasuyama (NITech/NIMS/JST) IBISML2018-1 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Level set estimation is a problem to identify a level set of an unknown function, which is defined by whether the function values are larger or smaller than a certain threshold. In practice, low cost approximate functions to the true function are often available. In this paper, we model the unknown true function and their approximate functions by multi-fidelity Gaussian process regression, based on which we propose an entropy-based active level set estimation method for cost-effective exploration. We demonstrate efficiency of our approach based on a parameter exploration problem in material science. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Gaussian Process / Multi-fidelity / Level Set Estimation / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 118, no. 81, IBISML2018-1, pp. 1-8, June 2018. |
Paper # |
IBISML2018-1 |
Date of Issue |
2018-06-06 (IBISML) |
ISSN |
Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
IBISML2018-1 |
|