Multi-objective optimization/bias-variance

[Antonio de Padua Braga]

Dear Connectionists,

The following paper has just been published by Neurocomputing.

The idea of the paper is to balance the error of the training
set and the norm of the weight vectors with a multi-objective
optimization approach to avoid over-fitting.

Copies are available on request.

We apologize in advance for any multiple postings that may be 
received.

***********************************************************************
Improving generalization of MLPs with multi-objective optimization

Teixeira, R.A., Braga, A.P., Takahashi, R.H.C. And Rezende, R. 
Neurocomputing. Volume 35, pages 189-194.

ABSTRACT

This paper presents a new learning scheme for improving 
generalization of Multilayer Perceptrons (MLPs). The algorithm 
uses a multi-objective optimization approach to balance between 
the error of the training data and  the  norm of network weight 
vectors to avoid over-fitting. The results are compared with
Support Vector Machines (SVMs) and standard backpropagation.

***********************************************************************
                 

-- 
Prof. Antonio de Padua Braga, Depto. Eng. Eletronica, Campus da
UFMG (Pampulha), C. P. 209, 30.161-970, Belo Horizonte, MG, Brazil
Tel:+55 31 4994869, Fax:+55 31 4994850, Email:apbraga at cpdee.ufmg.br,
http://www.cpdee.ufmg.br/~apbraga