A Novel Bayesian Logistic Discriminant Model with Dirichlet Distributions: An Application to Face Recognition

Authors:
Riadh Ksantini;Boubaker Boufama
Affiliations:
School of Computer Science, University of Windsor, Windsor, Canada N9B 3P4;School of Computer Science, University of Windsor, Windsor, Canada N9B 3P4
Venue:
ICIAR '09 Proceedings of the 6th International Conference on Image Analysis and Recognition
Year:
2009

Citing 8
Cited 0

Eigenfaces vs. Fisherfaces: Recognition Using Class Specific Linear Projection

IEEE Transactions on Pattern Analysis and Machine Intelligence
An introduction to support Vector Machines: and other kernel-based learning methods

An introduction to support Vector Machines: and other kernel-based learning methods
A Comparison of Prediction Accuracy, Complexity, and Training Time of Thirty-Three Old and New Classification Algorithms

Machine Learning
Bayesian parameter estimation via variational methods

Statistics and Computing
Empirical Performance Analysis of Linear Discriminant Classifiers

CVPR '98 Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
A deformable model for the recognition of human faces under arbitrary illumination

A deformable model for the recognition of human faces under arbitrary illumination
Weighted Pseudometric Discriminatory Power Improvement Using a Bayesian Logistic Regression Model Based on a Variational Method

IEEE Transactions on Pattern Analysis and Machine Intelligence
Unsupervised learning of a finite mixture model based on the Dirichlet distribution and its application

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

The Linear Discriminant Analysis (LDA) is a linear classifier which has proven to be powerful and competitive compared to the main state-of-the-art classifiers. However, the LDA assumes that the class conditional distributions are symmetric Gaussians with identical covariance structures, assumptions that are untrue for many classification and pattern recognition applications using heteroscedastic and asymmetric data. In this paper, a novel Bayesian Logistic Discriminant model with Dirichlet distributions (BLDD) is proposed to further relax the assumptions of the LDA by representing each class by a different Dirichlet distribution. At the same time, the BLDD tackles the so-called small sample size problem using a sparsity-promoting Gaussian prior over the unknown parameters. An extensive comparison of the BLDD to both LDA and Support Vector Machine (SVM) classifiers, performed on artificial and real datasets, has shown the advantages and superiority of our proposed method. In particular, the experiments on face recognition have clearly shown a significant improvement of the BLDD over the LDA.