講演抄録/キーワード |
講演名 |
2016-11-16 15:00
Robust supervised learning under uncertainty in dataset shift ○Weihua Hu・Issei Sato(UTokyo)・Masashi Sugiyama(RIKEN/UTokyo) IBISML2016-50 |
抄録 |
(和) |
(まだ登録されていません) |
(英) |
When machine learning is deployed in the real world, its performance can be significantly undermined because test data may follow a different distribution from training data. To build a reliable machine learning system in such a scenario, we propose a supervised learning framework that is explicitly robust to the uncertainty of dataset shift. Our robust learning framework is flexible in modeling various dataset shift scenarios. It is also computation- ally efficient in that it acts as a robust wrapper around existing gradient-based supervised learning algorithms, while adding negligible computational overheads. We discuss practical considerations in robust supervised learning and show the effectiveness of our approach on both synthetic and benchmark datasets. |
キーワード |
(和) |
/ / / / / / / |
(英) |
Dataset shift / Supervised learning / Robust learning / / / / / |
文献情報 |
信学技報, vol. 116, no. 300, IBISML2016-50, pp. 37-44, 2016年11月. |
資料番号 |
IBISML2016-50 |
発行日 |
2016-11-09 (IBISML) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
IBISML2016-50 |