IEICE Technical Committee Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2017-11-09 13:00
Consequently Fair Contextual Bandit Learning
Kazuto Fukuchi (Univ. of Tsukuba), Jun Sakuma (Univ. of Tsukuba/JST/RIKEN) IBISML2017-53
Abstract (in Japanese) (See Japanese page) 
(in English) Fairness in machine learning is being recognized as an important field. It requires that the consequent decisions made by a machine learning algorithm must not be biased on the sensitive attributes of individuals. In this paper, we introduce a novel notion of fairness, consequential fairness, and deal with sequential decision-making problems with a consequential fairness constraint. Consequential fairness requires almost sure satisfaction of fairness, that is, the number of unfair decisions does not exceed a prescribed threshold, $eta$, almost surely after a finite number of decisions. We first show a impossibility result for the adversarial context, which implies fundamental hardness of the setting. Meaningful results are obtained under the allocative contexts. We show that a novel algorithm, FairUcb, can achieve $O(max{ln T,min{eta,T/eta}})$ regret under a mild assumption. We also show that this regret bound is optimal up to constant if $eta le o(ln T)$.
Keyword (in Japanese) (See Japanese page) 
(in English) fairness / discrimination / contextual bandits / gap-dependent regret / adversarial contexts / allocative contexts / /  
Reference Info. IEICE Tech. Rep., vol. 117, no. 293, IBISML2017-53, pp. 139-146, Nov. 2017.
Paper # IBISML2017-53 
Date of Issue 2017-11-02 (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 IBISML2017-53

Conference Information
Committee IBISML  
Conference Date 2017-11-08 - 2017-11-10 
Place (in Japanese) (See Japanese page) 
Place (in English) Univ. of Tokyo 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Information-Based Induction Science Workshop (IBIS2017) 
Paper Information
Registration To IBISML 
Conference Code 2017-11-IBISML 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Consequently Fair Contextual Bandit Learning 
Sub Title (in English)  
Keyword(1) fairness  
Keyword(2) discrimination  
Keyword(3) contextual bandits  
Keyword(4) gap-dependent regret  
Keyword(5) adversarial contexts  
Keyword(6) allocative contexts  
Keyword(7)  
Keyword(8)  
1st Author's Name Kazuto Fukuchi  
1st Author's Affiliation University of Tsukuba (Univ. of Tsukuba)
2nd Author's Name Jun Sakuma  
2nd Author's Affiliation University of Tsukuba/Japan Science and Technology Agency CREST/RIKEN AIP (Univ. of Tsukuba/JST/RIKEN)
3rd Author's Name  
3rd Author's Affiliation ()
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
Speaker Author-1 
Date Time 2017-11-09 13:00:00 
Presentation Time 150 minutes 
Registration for IBISML 
Paper # IBISML2017-53 
Volume (vol) vol.117 
Number (no) no.293 
Page pp.139-146 
#Pages
Date of Issue 2017-11-02 (IBISML) 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan