This paper presents a new learning scheme for improving generalization of m
ultilayer perceptrons. The algorithm uses a multi-objective optimization ap
proach to balance between the error of the training data and the norm of ne
twork weight vectors to avoid overfitting. The results are compared with su
pport vector machines and standard backpropagation. (C) 2000 Elsevier Scien
ce B.V. All rights reserved.