A case study of a distributed high-performance computing system for neurocomputing

We model here a distributed implementation of cross-stopping, a combination of cross-validation and early-stopping techniques, for the selection of th e optimal architecture of feed-forward networks. Due to the very large comp utational demand of the method, we use the RAIN system (Redundant Array of Inexpensive workstations for Neurocomputing) as a target platform for the e xperiments and show that this kind of system can be effectively used for co mputational intensive neurocomputing tasks. (C) 2000 Elsevier Science B.V. All rights reserved.