Multiple view semi-supervised dimensionality reduction

Multiple view data, together with some domain knowledge in the form of pairwise constraints, arise in various data mining applications. How to learn a hidden consensus pattern in the low dimensional space is a challenging problem. In this paper, we propose a new method for multiple view semi-supervi...

Full description

Saved in:
Bibliographic Details
Published in:Pattern recognition 2010-03, Vol.43 (3), p.720-730
Main Authors: Hou, Chenping, Zhang, Changshui, Wu, Yi, Nie, Feiping
Format: Article
Language:English
Subjects:
Applied sciences
Dimensionality reduction
Domain knowledge
Exact sciences and technology
Information theory
Information, signal and communications theory
Multiple view
Semi-supervised
Telecommunications and information theory
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Multiple view data, together with some domain knowledge in the form of pairwise constraints, arise in various data mining applications. How to learn a hidden consensus pattern in the low dimensional space is a challenging problem. In this paper, we propose a new method for multiple view semi-supervised dimensionality reduction. The pairwise constraints are used to derive embedding in each view and simultaneously, the linear transformation is introduced to make different embeddings from different pattern spaces comparable. Hence, the consensus pattern can be learned from multiple embeddings of multiple representations. We derive an iterating algorithm to solve the above problem. Some theoretical analyses and out-of-sample extensions are also provided. Promising experiments on various data sets, together with some important discussions, are also presented to demonstrate the effectiveness of the proposed algorithm.
ISSN:0031-3203
1873-5142
DOI:10.1016/j.patcog.2009.07.015