Paper Abstract and Keywords |
Presentation |
2016-09-06 10:45
A proposal on quick sensitivity analysis of empirical risk minimization problems Hiroyuki Hanada, Atsushi Shibagaki (NITech), Jun Sakuma (Univ. of Tsukuba), Ichiro Takeuchi (NITech) PRMU2016-80 IBISML2016-35 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
For a training data set consisting of $n$ vectors of $d$ dimensions, we consider obtaining a training result from it by a regularized empirical risk minimization. In such a situation, if we need a new training result when a part of the training data, we need $O(nd)$ time in general even if we know the training result before the change of the data. A solution for the problem is the quick sensitivity analysis, which computes the upper and the lower bounds of the training result rather than the exact one with much smaller computational cost.
There has been a quick sensitivity analysis method for instance-wise change, which enables us to compute the upper and the lower bounds of the training result in $O(d|{cal M}_I|)$ time for the number of changed instances $|{cal M}_I|$ ($leq n$). On the other hand, we propose a fast method for more generalized data changes: it enables us to compute the upper and the lower bounds of the training result in $O(d|{cal M}|)$ time for the number of changed values $|{cal M}|$ ($leq nd$). |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Empirical risk minimization / Quick sensitivity analysis / Strong convexity / Duality gap / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 116, no. 209, IBISML2016-35, pp. 203-210, Sept. 2016. |
Paper # |
IBISML2016-35 |
Date of Issue |
2016-08-29 (PRMU, 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 |
PRMU2016-80 IBISML2016-35 |
Conference Information |
Committee |
PRMU IPSJ-CVIM IBISML |
Conference Date |
2016-09-05 - 2016-09-06 |
Place (in Japanese) |
(See Japanese page) |
Place (in English) |
|
Topics (in Japanese) |
(See Japanese page) |
Topics (in English) |
|
Paper Information |
Registration To |
IBISML |
Conference Code |
2016-09-PRMU-CVIM-IBISML |
Language |
Japanese |
Title (in Japanese) |
(See Japanese page) |
Sub Title (in Japanese) |
(See Japanese page) |
Title (in English) |
A proposal on quick sensitivity analysis of empirical risk minimization problems |
Sub Title (in English) |
|
Keyword(1) |
Empirical risk minimization |
Keyword(2) |
Quick sensitivity analysis |
Keyword(3) |
Strong convexity |
Keyword(4) |
Duality gap |
Keyword(5) |
|
Keyword(6) |
|
Keyword(7) |
|
Keyword(8) |
|
1st Author's Name |
Hiroyuki Hanada |
1st Author's Affiliation |
Nagoya Institute of Technology (NITech) |
2nd Author's Name |
Atsushi Shibagaki |
2nd Author's Affiliation |
Nagoya Institute of Technology (NITech) |
3rd Author's Name |
Jun Sakuma |
3rd Author's Affiliation |
University of Tsukuba (Univ. of Tsukuba) |
4th Author's Name |
Ichiro Takeuchi |
4th Author's Affiliation |
Nagoya Institute of Technology (NITech) |
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 |
2016-09-06 10:45:00 |
Presentation Time |
30 minutes |
Registration for |
IBISML |
Paper # |
PRMU2016-80, IBISML2016-35 |
Volume (vol) |
vol.116 |
Number (no) |
no.208(PRMU), no.209(IBISML) |
Page |
pp.203-210 |
#Pages |
8 |
Date of Issue |
2016-08-29 (PRMU, IBISML) |
|