A METHOD FOR IMPROVING CLASSIFICATION RELIABILITY OF MULTILAYER PERCEPTRONS

Criteria for evaluating the classification reliability of a neural cla ssifier and for accordingly making a reject option are proposed, Such an option, implemented by means of two rules which can be applied inde pendently of topology, size, and training algorithms of the neural cla ssifier, allows to improve the classification reliability, It is assum ed that a performance function P is defined which, taking into account the requirements of the particular application, evaluates the quality of the classification in terms of recognition, misclassification, and reject rates, Under this assumption the optimal reject threshold valu e, determining the best trade-off between reject rate and misclassific ation rate, is the one for which the function P reaches its absolute m aximum, No constraints are imposed on the form of P, but the ones nece ssary in order that P actually measures the quality of the classificat ion process, The reject threshold is evaluated on the basis of some st atistical distributions characterizing the behavior of the classifier when operating without reject option; these distributions are computed once the training phase of the net has been completed, The method has been tested with a neural classifier devised for handprinted and mult ifont printed characters, by using a database of about 300000 samples. Experimental results are discussed.