The most common application of neural networks in cognitive psychology is to implement formal models of information processing in the human cognitive system. Thus, they are used in the process of testing cognitive theories. A more pragmatic use of these neuro-computational systems concerns tasks of classification and prediction. Because of the mathematical models they use, neural networks reach significantly higher performances in tasks of prediction and classification, compared to the usual statistical methods (Ripley, 1993; 1994). All the more frequently, the networks are used as a device for statistical calculus. We first present in this paper the use of neural networks in a classification task. Several patients suffering from different psychopathological disorders are classified based on their scores at YSQ (Young Schema Questionnaire - Young, 1990). The best neural network we found has an excellent performance, classifying without errors the patients according to their diagnosis. We next use an RBF neural network to accomplish a prediction task, using the patient's scores at YSQ as an independent variable and the disorders they have been diagnosed with as dependent one. From a sample of 140 subjects, the network is first trained using the YSQ scores from 100 of them. The other 40 are used for testing the network. We test if the network can predict, based on the patterns it has learned in the training stage, the psychopathological categories to which the other 40 subjects (whose data have not been presented to the network in the training stage) belong to. The network's performance in this testing stage is 92,5%, a percent much superior to the ones obtained by any classical regression model.
Keywords: neural networks, prediction, classification, psychopathology