Exploiting quadratic mutual information for discriminant analysis

Authors:
Vasileios Gavriilidis;Anastasios Tefas
Affiliations:
Department of Informatics, Aristotle University of Thessaloniki, Thessaloniki, Greece;Department of Informatics, Aristotle University of Thessaloniki, Thessaloniki, Greece
Venue:
SETN'12 Proceedings of the 7th Hellenic conference on Artificial Intelligence: theories and applications
Year:
2012

Citing 8
Cited 0

Introduction to statistical pattern recognition (2nd ed.)

Introduction to statistical pattern recognition (2nd ed.)
Independent component analysis, a new concept?

Signal Processing - Special issue on higher order statistics
Approximate statistical tests for comparing supervised classification learning algorithms

Neural Computation
Heteroscedastic discriminant analysis and reduced rank HMMs for improved speech recognition

Speech Communication
Feature extraction by non parametric mutual information maximization

The Journal of Machine Learning Research
Spectral feature projections that maximize Shannon mutual information with class labels

Pattern Recognition
Linear projection method based on information theoretic learning

ICANN'10 Proceedings of the 20th international conference on Artificial neural networks: Part III
Statistical Properties of Error Estimators in Performance Assessment of Recognition Systems

IEEE Transactions on Pattern Analysis and Machine Intelligence

Quantified Score

Hi-index	0.00

Visualization

Abstract

Novel criteria that reformulate the Quadratic Mutual Information according to Fisher's Discriminant Analysis are proposed for supervised dimensionality reduction. The proposed method uses a quadratic divergence measure and requires no prior assumptions about class densities. The criteria are optimized using gradient ascent with initialization using random or LDA based projections. Experiments on various datasets are conducted and highlight the superiority of the proposed approach compared to the standard QMI criterion.