CaltechAUTHORS
  A Caltech Library Service

Efficient Large-Scale Structured Learning

Branson, Steve and Beijbom, Oscar and Belongie, Serge (2013) Efficient Large-Scale Structured Learning. In: 2013 Computer Vision and Pattern Recognition (CVPR). IEEE Conference on Computer Vision and Pattern Recognition. IEEE , New York, NY, pp. 1806-1813. ISBN 978-0-7695-4989-7. https://resolver.caltech.edu/CaltechAUTHORS:20140324-113612849

Full text is not posted in this repository. Consult Related URLs below.

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20140324-113612849

Abstract

We introduce an algorithm, SVM-IS, for structured SVM learning that is computationally scalable to very large datasets and complex structural representations. We show that structured learning is at least as fast-and often much faster-than methods based on binary classification for problems such as deformable part models, object detection, and multiclass classification, while achieving accuracies that are at least as good. Our method allows problem-specific structural knowledge to be exploited for faster optimization by integrating with a user-defined importance sampling function. We demonstrate fast train times on two challenging large scale datasets for two very different problems: Image Net for multiclass classification and CUB-200-2011 for deformable part model training. Our method is shown to be 10-50 times faster than SVM^(struct) for cost-sensitive multiclass classification while being about as fast as the fastest 1-vs-all methods for multiclass classification. For deformable part model training, it is shown to be 50-1000 times faster than methods based on SVM^(struct), mining hard negatives, and Pegasos-style stochastic gradient descent. Source code of our method is publicly available.


Item Type:Book Section
Related URLs:
URLURL TypeDescription
http://dx.doi.org/10.1109/CVPR.2013.236DOIArticle
http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6619080PublisherArticle
Additional Information:© 2013 IEEE. Date of Conference: 23-28 June 2013. The authors thank Kai Wang and Catherine Wah for helpful discussions and feedback. Funding for this work was provided by NSF Computer Vision Coral Ecology grant #ATM-0941760 and the Amazon AWS in Education program. INSPEC Accession Number:13824469.
Funders:
Funding AgencyGrant Number
NSF Computer Vision Coral EcologyATM-0941760
Amazon AWS in Education program UNSPECIFIED
Series Name:IEEE Conference on Computer Vision and Pattern Recognition
Record Number:CaltechAUTHORS:20140324-113612849
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20140324-113612849
Official Citation:Branson, S.; Beijbom, O.; Belongie, S., "Efficient Large-Scale Structured Learning," Computer Vision and Pattern Recognition (CVPR), 2013 IEEE Conference on , vol., no., pp.1806,1813, 23-28 June 2013 doi: 10.1109/CVPR.2013.236
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:44461
Collection:CaltechAUTHORS
Deposited By: Ruth Sustaita
Deposited On:26 Mar 2014 21:17
Last Modified:03 Oct 2019 06:18

Repository Staff Only: item control page