Bayesian recognition improved by automatic associations of variables

Abstract

A rather complex improvement of Bayes formula is presented. Although it is slow in computation, experimental results sugges that its range of validity is double that of standard methods when the given symptoms do not satisfy the assumption of independency (which is needed to apply Bayes formula to pattern recognition problems in the presence of many symptoms).

^†Lavoro svolto nellambito del santratto 115.3050.0.5189 del C.N R. presso l'Univeisita di Genova (Italy).

^†Lavoro svolto nellambito del santratto 115.3050.0.5189 del C.N R. presso l'Univeisita di Genova (Italy).

Notes

^†Lavoro svolto nellambito del santratto 115.3050.0.5189 del C.N R. presso l'Univeisita di Genova (Italy).