A competitive model for semi-supervised discriminant analysis

Authors:
Weifu Chen;Guocan Feng;Xiaolin Zou;Zhiyong Liu
Affiliations:
School of Math. and Comput. Science, Sun Yat-sen University, Guangzhou, China;School of Math. and Comput. Science, Sun Yat-sen University, Guangzhou, China;School of Math. and Inform. Science, Zhaoqing University, Zhaoqing, China;School of Math. and Comput. Science, Sun Yat-sen University, Guangzhou, China, Industry Center, ShenZhen Polytechnic, Shenzhen, China
Venue:
CCBR'12 Proceedings of the 7th Chinese conference on Biometric Recognition
Year:
2012

Citing 8
Cited 0

Pattern Classification (2nd Edition)

Pattern Classification (2nd Edition)
Discriminative Common Vectors for Face Recognition

IEEE Transactions on Pattern Analysis and Machine Intelligence
Manifold Regularization: A Geometric Framework for Learning from Labeled and Unlabeled Examples

The Journal of Machine Learning Research
A study on three linear discriminant analysis based methods in small sample size problem

Pattern Recognition
A unified framework for semi-supervised dimensionality reduction

Pattern Recognition
Discriminability and reliability indexes: Two new measures to enhance multi-image face recognition

Pattern Recognition
Flexible manifold embedding: a framework for semi-supervised and unsupervised dimension reduction

IEEE Transactions on Image Processing
Efficient and robust feature extraction by maximum margin criterion

IEEE Transactions on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a general linear framework and a competitive model for discriminant analysis with partially labeled data. Our method first utilizes the competitive model to find the reliable training samples. Two indices are given to measure the reliability. In the second stage, discriminant vectors are computed by the proposed framework. We show that under different graph models some popular discriminant analysis algorithms are special cases of the proposed framework. Experimental results suggest that our algorithm is effective and can significantly improve the recognition accuracy.