Self-organized classification problem solving with yprel neural networks

This paper deals with a new scheme of distributed classifier based on a particular formal neuron named “yprel”. The main characteristics of the proposed approach are: (i) a classifier is a set of interconnected and cooperating networks, (ii) the distributed resolution strategy emerges from the individual network classification behaviors during the incremental building phase of the classifier; (iii) each neuron is able to come to classification decisions about some elements and to communicate them; (iv) the network architectures and the interconnexion links between the networks are not a priori chosen, but get themselves organized thanks to an incremental and competitive learning between the decision-making neurons