Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold

Learning robust subspaces to maximize class discrimination is challenging, and most current works consider a weak connection between dimensionality reduction and classifier design. We propose an alternate framework wherein these two steps are combined in a joint formulation to exploit the direct con...

Full description

Bibliographic Details
Main Authors: Pham, DucSon, Venkatesh, Svetha
Other Authors: D Wang
Format: Conference Paper
Published: Springer 2011
Online Access:http://hdl.handle.net/20.500.11937/42511
_version_ 1848756439535845376
author Pham, DucSon
Venkatesh, Svetha
author2 D Wang
author_facet D Wang
Pham, DucSon
Venkatesh, Svetha
author_sort Pham, DucSon
building Curtin Institutional Repository
collection Online Access
description Learning robust subspaces to maximize class discrimination is challenging, and most current works consider a weak connection between dimensionality reduction and classifier design. We propose an alternate framework wherein these two steps are combined in a joint formulation to exploit the direct connection between dimensionality reduction and classification. Specifically, we learn an optimal subspace on the Grassmann manifold jointly minimizing the classification error of an SVM classifier. We minimize the regularized empirical risk over both the hypothesis space of functions that underlies this new generalized multiclass Lagrangian SVM and the Grassmann manifold such that a linear projection is to be found. We propose an iterative algorithm to meet the dual goal of optimizing both the classifier and projection. Extensive numerical studies on challenging datasets show robust performance of the proposed scheme over other alternatives in contexts wherein limited training data is used, verifying the advantage of the joint formulation.
first_indexed 2025-11-14T09:12:13Z
format Conference Paper
id curtin-20.500.11937-42511
institution Curtin University Malaysia
institution_category Local University
last_indexed 2025-11-14T09:12:13Z
publishDate 2011
publisher Springer
recordtype eprints
repository_type Digital Repository
spelling curtin-20.500.11937-425112023-01-27T05:52:12Z Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold Pham, DucSon Venkatesh, Svetha D Wang M Reynolds Learning robust subspaces to maximize class discrimination is challenging, and most current works consider a weak connection between dimensionality reduction and classifier design. We propose an alternate framework wherein these two steps are combined in a joint formulation to exploit the direct connection between dimensionality reduction and classification. Specifically, we learn an optimal subspace on the Grassmann manifold jointly minimizing the classification error of an SVM classifier. We minimize the regularized empirical risk over both the hypothesis space of functions that underlies this new generalized multiclass Lagrangian SVM and the Grassmann manifold such that a linear projection is to be found. We propose an iterative algorithm to meet the dual goal of optimizing both the classifier and projection. Extensive numerical studies on challenging datasets show robust performance of the proposed scheme over other alternatives in contexts wherein limited training data is used, verifying the advantage of the joint formulation. 2011 Conference Paper http://hdl.handle.net/20.500.11937/42511 10.1007/978-3-642-25832-9_25 Springer restricted
spellingShingle Pham, DucSon
Venkatesh, Svetha
Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold
title Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold
title_full Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold
title_fullStr Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold
title_full_unstemmed Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold
title_short Supervised subspace learning with multi-class Lagrangian SVM on the Grassmann Manifold
title_sort supervised subspace learning with multi-class lagrangian svm on the grassmann manifold
url http://hdl.handle.net/20.500.11937/42511