摘要：In this article, we investigated the machine learning based on the mathematics modeling bayesian and connectionist, applied to support in the diagnostic of epileptic events (EEs) and non-epileptic events (NEEs). To this end, some algorithms of machine learning were compared including the connectionist learning through Generalized Delta, Hebb and Oja rules learning, with the learning in Bayesian Networks. There were considered to this study 122 patients which showed symptoms related to EEs and NEEs. The empiric results of this search indicate that: (1) both bayesian and connectionist approaches are too similar, revealing consistent results, even using distinct heuristics; (2) the indices of positive cases (EEs), from the bayesian modeling may be related to the facility of the implementation of the conditional probabilities. It was concluded, therefore that the investigated models give important information to the development of new searches as in the bayesian field as in the connectionist field.
