Paper Abstract and Keywords |
Presentation |
2013-07-18 13:50
An improvement of a Neutrality Term in an Information-neutral Recommender System Toshihiro Kamishima, Shotaro Akaho, Hideki Asoh (AIST), Jun Sakuma (Univ. of Tsukuba) IBISML2013-7 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Information-neutral recommender systems aim to make recommendations whose neutrality from the specified viewpoint is guaranteed. Such systems is developed for dissolving a filter bubble problem, which is the bias or restriction that provided to people by the influence of personalization technologies. Our previously developed system was not scalable because efficient optimization techniques could not be applied. To address this problem, we developed a penalty term to guarantee the neutrality that can be analytically differentiable. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
fairness-aware data mining / recommender system / matrix factorization / neutrality / filter bubble / / / |
Reference Info. |
IEICE Tech. Rep., vol. 113, no. 139, IBISML2013-7, pp. 43-50, July 2013. |
Paper # |
IBISML2013-7 |
Date of Issue |
2013-07-11 (IBISML) |
ISSN |
Print edition: ISSN 0913-5685 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 |
IBISML2013-7 |
|