Bregman Divergence-Based Regularization for Transfer Subspace Learning

The regularization principals [31] lead approximation schemes to deal with various learning problems, e.g., the regularization of the norm in a reproducing kernel Hilbert space for the ill-posed problem. In this paper, we present a family of subspace learning algorithms based on a new form of regula...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Veröffentlicht in:	IEEE transactions on knowledge and data engineering Jg. 22; H. 7; S. 929 - 942
Hauptverfasser:	Si, Si, Tao, Dacheng, Geng, Bo
Format:	Journal Article
Sprache:	Englisch
Veröffentlicht:	New York, NY IEEE 01.07.2010 IEEE Computer Society The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:	Algorithm design and analysis Algorithms and Bregman divergence Applied sciences Artificial intelligence Computer science; control theory; systems Data processing. List processing. Character string processing Dimensionality reduction Discriminant analysis Exact sciences and technology Face recognition Hilbert space Independent component analysis Information systems. Data bases Kernel Learning Linear discriminant analysis Memory organisation. Data processing Pattern recognition. Digital image processing. Computational geometry Principal component analysis Regularization Semisupervised learning Software Speech and sound recognition and synthesis. Linguistics Studies Subspaces Testing Text categorization Texts Training Image processing regularization Modeling Learning Alignment Vector space Classification Facies Hilbert space Ill posed problem Learning algorithm Dimensionality reduction Discriminant analysis Face recognition Locality Image databank Text Pattern recognition Character recognition Knowledge transfer Fisher information Dimension reduction and Bregman divergence Artificial intelligence Principal component analysis
ISSN:	1041-4347, 1558-2191
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Beschreibung
Zusammenfassung:	The regularization principals [31] lead approximation schemes to deal with various learning problems, e.g., the regularization of the norm in a reproducing kernel Hilbert space for the ill-posed problem. In this paper, we present a family of subspace learning algorithms based on a new form of regularization, which transfers the knowledge gained in training samples to testing samples. In particular, the new regularization minimizes the Bregman divergence between the distribution of training samples and that of testing samples in the selected subspace, so it boosts the performance when training and testing samples are not independent and identically distributed. To test the effectiveness of the proposed regularization, we introduce it to popular subspace learning algorithms, e.g., principal components analysis (PCA) for cross-domain face modeling; and Fisher's linear discriminant analysis (FLDA), locality preserving projections (LPP), marginal Fisher's analysis (MFA), and discriminative locality alignment (DLA) for cross-domain face recognition and text categorization. Finally, we present experimental evidence on both face image data sets and text data sets, suggesting that the proposed Bregman divergence-based regularization is effective to deal with cross-domain learning problems.
Bibliographie:	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14 ObjectType-Article-2 ObjectType-Feature-1 content type line 23
ISSN:	1041-4347 1558-2191
DOI:	10.1109/TKDE.2009.126