ML-MG: Multi-label Learning with Missing Labels Using a Mixed Graph

Handle URI:
http://hdl.handle.net/10754/605068
Title:
ML-MG: Multi-label Learning with Missing Labels Using a Mixed Graph
Authors:
Wu, Baoyuan ( 0000-0003-2183-5990 ) ; Lyu, Siwei; Ghanem, Bernard ( 0000-0002-5534-587X )
Abstract:
This work focuses on the problem of multi-label learning with missing labels (MLML), which aims to label each test instance with multiple class labels given training instances that have an incomplete/partial set of these labels (i.e. some of their labels are missing). To handle missing labels, we propose a unified model of label dependencies by constructing a mixed graph, which jointly incorporates (i) instance-level similarity and class co-occurrence as undirected edges and (ii) semantic label hierarchy as directed edges. Unlike most MLML methods, We formulate this learning problem transductively as a convex quadratic matrix optimization problem that encourages training label consistency and encodes both types of label dependencies (i.e. undirected and directed edges) using quadratic terms and hard linear constraints. The alternating direction method of multipliers (ADMM) can be used to exactly and efficiently solve this problem. To evaluate our proposed method, we consider two popular applications (image and video annotation), where the label hierarchy can be derived from Wordnet. Experimental results show that our method achieves a significant improvement over state-of-the-art methods in performance and robustness to missing labels.
KAUST Department:
Visual Computing Center (VCC); Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
Publisher:
Institute of Electrical and Electronics Engineers (IEEE)
Journal:
2015 IEEE International Conference on Computer Vision (ICCV)
Conference/Event name:
2015 IEEE International Conference on Computer Vision (ICCV)
Issue Date:
7-Dec-2015
DOI:
10.1109/ICCV.2015.473
Type:
Conference Paper
Sponsors:
This work is supported supported by competitive research funding from King Abdullah University of Science and Technology (KAUST). The participation of Siwei Lyu in this work is partly supported by US National Science Foundation Research Grant (CCF-1319800) and National Science Foundation Early Faculty Career Development (CAREER) Award (IIS-0953373). We thank Fabian Caba Heilbron for his help on figure plotting, and Rafal Protasiuk for his help on data collection. We thank the reviewers for their constructive comments.
Additional Links:
http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7410830
Appears in Collections:
Conference Papers; Visual Computing Center (VCC); Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division

Full metadata record

DC FieldValue Language
dc.contributor.authorWu, Baoyuanen
dc.contributor.authorLyu, Siweien
dc.contributor.authorGhanem, Bernarden
dc.date.accessioned2016-04-12T09:40:03Zen
dc.date.available2016-04-12T09:40:03Zen
dc.date.issued2015-12-07en
dc.identifier.doi10.1109/ICCV.2015.473en
dc.identifier.urihttp://hdl.handle.net/10754/605068en
dc.description.abstractThis work focuses on the problem of multi-label learning with missing labels (MLML), which aims to label each test instance with multiple class labels given training instances that have an incomplete/partial set of these labels (i.e. some of their labels are missing). To handle missing labels, we propose a unified model of label dependencies by constructing a mixed graph, which jointly incorporates (i) instance-level similarity and class co-occurrence as undirected edges and (ii) semantic label hierarchy as directed edges. Unlike most MLML methods, We formulate this learning problem transductively as a convex quadratic matrix optimization problem that encourages training label consistency and encodes both types of label dependencies (i.e. undirected and directed edges) using quadratic terms and hard linear constraints. The alternating direction method of multipliers (ADMM) can be used to exactly and efficiently solve this problem. To evaluate our proposed method, we consider two popular applications (image and video annotation), where the label hierarchy can be derived from Wordnet. Experimental results show that our method achieves a significant improvement over state-of-the-art methods in performance and robustness to missing labels.en
dc.description.sponsorshipThis work is supported supported by competitive research funding from King Abdullah University of Science and Technology (KAUST). The participation of Siwei Lyu in this work is partly supported by US National Science Foundation Research Grant (CCF-1319800) and National Science Foundation Early Faculty Career Development (CAREER) Award (IIS-0953373). We thank Fabian Caba Heilbron for his help on figure plotting, and Rafal Protasiuk for his help on data collection. We thank the reviewers for their constructive comments.en
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)en
dc.relation.urlhttp://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7410830en
dc.rights(c) 2015 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material for advertising or promotional purposes, creating new collective works for resale or redistribution to servers or lists, or reuse of any copyrighted components of this work in other works.en
dc.titleML-MG: Multi-label Learning with Missing Labels Using a Mixed Graphen
dc.typeConference Paperen
dc.contributor.departmentVisual Computing Center (VCC)en
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Divisionen
dc.identifier.journal2015 IEEE International Conference on Computer Vision (ICCV)en
dc.conference.date7-13 Dec. 2015en
dc.conference.name2015 IEEE International Conference on Computer Vision (ICCV)en
dc.conference.locationSantiagoen
dc.eprint.versionPost-printen
dc.contributor.institutionUniversity at Albany, SUNY Albany, NY, USAen
kaust.authorWu, Baoyuanen
kaust.authorGhanem, Bernarden
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.