Problem specific network structure optimization subsumes the problem of inp
ut selection and network topology identification. Requirements to the netwo
rk should be accuracy and good generalization abilities. In this contributi
on we describe in detail an evolutionary algorithm which performs both task
s well. Furthermore, approximation results on mathematical and real world d
ata are presented. In this case we used lattice-based associative memory ne
tworks (LB-AMNs) using B-splines as basis functions. The method here is not
restricted to B-splines as basis functions. The proposed method and algori
thm can be seen as optimized classification system. (C) 2001 Published by E
lsevier Science Inc.