MPI-INF Logo
Publications

Server    domino.mpi-inf.mpg.de

Proceedings Article, Paper
@InProceedings
Beitrag in Tagungsband, Workshop

Author, Editor
Author(s):
Ebert, Sandra
Fritz, Mario
Schiele, Bernt		dblp
dblp
dblp
Editor(s):
Belongie, Serge
Blake, Andrew
Luo, Jiebo
Yuille, Alan		dblp
dblp
dblp
dblp
Not MPII Editor(s):
Belongie, Serge
Blake, Andrew
Luo, Jiebo
Yuille, Alan
BibTeX cite key*:
Ebert2012CVPR
Title, Booktitle
Title*:
RALF: A Reinforced Active Learning Formulation for Object Class Recognition
Booktitle*:
2012 IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2012)
Event, URLs
Conference URL::
http://www.cvpr2012.org/
Downloading URL:
Event Address*:
Providence, RI
Language:
English
Event Date*
(no longer used):
Organization:
Event Start Date:
18 June 2012
Event End Date:
20 June 2012
Publisher
Name*:
IEEE
URL:
Address*:
Piscataway, NJ
Type:
Vol, No, Year, pp.
Series:
Volume:
Number:
Month:
June
Pages:
3626-3633
Year*:
2012
VG Wort Pages:
ISBN/ISSN:
978-1-4673-1226-4
Sequence Number:
DOI:
10.1109/CVPR.2012.6248108
Note, Abstract, ©
(LaTeX) Abstract:
Active learning aims to reduce the amount of labels required for classification. The main difficulty is to find a
good trade-off between exploration and exploitation of the
labeling process that depends – among other things – on
the difficulty of the classification task, the distribution of
the data and the employed classification scheme. In this paper, we analyze different sampling criteria including a novel
density-based criteria and demonstrate the importance to
combine exploration and exploitation sampling criteria. We
also show that a time-varying combination of sampling criteria often improves performance. Finally, by formulating
the criteria selection as a Markov decision process, we
propose a novel feedback-driven framework based on reinforcement learning. This does not require prior information
on the dataset or the sampling criteria but rather is able to
adapt the sampling strategy during the learning process by
experience. We evaluate our approach on three challenging
object recognition datasets and show superior performance
to previous active learning methods.
Download
Access Level:
Public

Correlation
MPG Unit:
Max-Planck-Institut für Informatik
MPG Subunit:
Computer Vision and Multimodal Computing
Appearance:
MPII WWW Server, MPII FTP Server, MPG publications list, university publications list, working group publication list, Fachbeirat, VG Wort



BibTeX Entry:

@INPROCEEDINGS{Ebert2012CVPR,
AUTHOR = {Ebert, Sandra and Fritz, Mario and Schiele, Bernt},
 EDITOR = {Belongie, Serge and Blake, Andrew and Luo, Jiebo and Yuille, Alan},
 TITLE = {{RALF}: A Reinforced Active Learning Formulation for Object Class Recognition},
 BOOKTITLE = {2012 IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2012)},
 PUBLISHER = {IEEE},
 YEAR = {2012},
 PAGES = {3626--3633},
 ADDRESS = {Providence, RI},
 MONTH = {June},
 ISBN = {978-1-4673-1226-4},
 DOI = {10.1109/CVPR.2012.6248108},
}

Entry last modified by Anja Becker, 03/08/2013
Hide details for Edit History (please click the blue arrow to see the details)Edit History (please click the blue arrow to see the details)

Editor(s)
[Library]		Created
03/07/2012 10:43:24
Revisions
3.
2.
1.
0.
Editor(s)
Anja Becker
Anja Becker
Sandra Ebert
Sandra Ebert
Edit Dates
08.03.2013 11:11:30
08.02.2013 14:50:51
12/03/2012 04:18:35 PM
03/07/2012 10:43:24 AM