FUZZY PATTERN-CLASSIFICATION AND THE CONNECTIONIST APPROACH

Several hybrid architectures combining fuzzy pattern classification an d the connectionist approach will be developed and tested for the part icular problem of diagnostic classification in computerized electrocar diography. The first level of fuzzy description of the input parameter s is performed by a layer of Radial Basis Functions, and this step can be seen as a level of data abstraction. A subsequent classical NN pro cesses these fuzzy descriptions. Several experiments have been perform ed on the components of the resulting architecture in order to point o ut their influence on the overall performance in the diagnostic classi fication task. A large validated database has been used for the valida tion of the proposed hybrid architecture.