Community Research and Development Information Service - CORDIS


This report describes some tests which make reference to a neural networks simulator, BBackprop, to test the theory expounded in 1992. The simulator was written by MacKay in C and tested on Sun SparcStation. BBackprop integrates the functionalities of a typical neural networks simulator with other tools such as a criterion to decide among a set of possible models. these new functionalities are based on common methods of Bayesian statistics applied to neural networks for the first time by MacKay. BBackprop has been recompiled in different situations and compared the results obtained by executing the program on a set of benchmark problems.

Additional information

Authors: PELA B, JRC Ispra (IT);PERROTTA D, JRC Ispra (IT)
Bibliographic Reference: Article: Scienza e Supercalcolo al Cineca-Rapporto 95
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top