Linear projection method based on information theoretic learning

Authors:
Pablo A. Vera;Pablo A. Estévez;Jose C. Principe
Affiliations:
Department of Electrical Engineering and Advanced Mining Technology Center, Universidad de Chile, Santiago, Chile;Department of Electrical Engineering and Advanced Mining Technology Center, Universidad de Chile, Santiago, Chile;CNEL, University of Florida
Venue:
ICANN'10 Proceedings of the 20th international conference on Artificial neural networks: Part III
Year:
2010

Citing 9
Cited 1

An application of the principle of maximum information preservation to linear systems

Advances in neural information processing systems 1
Independent component analysis: algorithms and applications

Neural Networks
Feature extraction by non parametric mutual information maximization

The Journal of Machine Learning Research
VizRank: Data Visualization Guided by Machine Learning

Data Mining and Knowledge Discovery
Online data visualization using the neural gas network

Neural Networks - 2006 Special issue: Advances in self-organizing maps--WSOM'05
A Nonlinear Mapping for Data Structure Analysis

IEEE Transactions on Computers
Clustering

Clustering
Relevance measures for subset variable selection in regression problems based on k-additive mutual information

Computational Statistics & Data Analysis
Information Theoretic Learning: Renyi's Entropy and Kernel Perspectives

Information Theoretic Learning: Renyi's Entropy and Kernel Perspectives

Exploiting quadratic mutual information for discriminant analysis

SETN'12 Proceedings of the 7th Hellenic conference on Artificial Intelligence: theories and applications

Quantified Score

Hi-index	0.00

Visualization

Abstract

A new unsupervised method for linear projection of multidimensional data based on Linsker's principle of maximum information preservation is proposed. The Quadratic Mutual Information (QMI) between the input X and the output Y is estimated, assuming a linear mapping. This estimation is made using a non-parametric quadratic divergence measure, without any assumption of data distribution. The results show that the 2D projections obtained with the proposed method are better than PCA projections in terms of cluster separability.