Optimization of the kernel functions in a probabilistic neural network analyzing the local pattern distribution

Authors:
I. Galleske;J. Castellanos
Affiliations:
Departmento de Inteligencia Artificial, Facultad de Informática, Universidad Politécnica de Madrid, Spain;Departmento de Inteligencia Artificial, Facultad de Informática, Universidad Politécnica de Madrid, Spain
Venue:
Neural Computation
Year:
2002

Citing 4
Cited 7

Probabilistic neural networks

Neural Networks
Robust maximum likelihood training of heteroscedastic probabilistic neural networks

Neural Networks
On the Generalization Ability of Neural Network Classifiers

IEEE Transactions on Pattern Analysis and Machine Intelligence
Probabilistic Neural Networks with Rotated Kernel Functions

ICANN '97 Proceedings of the 7th International Conference on Artificial Neural Networks

Gradient-Based Adaptation of General Gaussian Kernels

Neural Computation
Similarity classifier using similarity measure derived from Yu's norms in classification of medical data sets

Computers in Biology and Medicine
Gap-Based Estimation: Choosing the Smoothing Parameters for Probabilistic and General Regression Neural Networks

Neural Computation
Classification method using fuzzy level set subgrouping

Expert Systems with Applications: An International Journal
Variations of the two-spiral task

Connection Science
On detecting nonlinear patterns in discriminant problems

Information Sciences: an International Journal
Supervised Learning Probabilistic Neural Networks

Neural Processing Letters

Quantified Score

Hi-index	0.00

Visualization

Abstract

This article proposes a procedure for the automatic determination of the elements of the covariance matrix of the gaussian kernel function of probabilistic neural networks. Two matrices, a rotation matrix and a matrix of variances, can be calculated by analyzing the local environment of each training pattern. The combination of them will form the covariance matrix of each training pattern. This automation has two advantages: First, it will free the neural network designer from indicating the complete covariance matrix, and second, it will result in a network with better generalization ability than the original model. A variation of the famous two-spiral problem and real-world examples from the UCI Machine Learning Repository will show a classification rate not only better than the original probabilistic neural network but also that this model can outperform other well-known classification techniques.