Modified counterpropagation employing neo fuzzy neurons and its application to system modeling
K. Horio, T. Yamakawa · 2002
In this paper, a modified counterpropagation employing neo fuzzy neurons is proposed. The counterpropagation is a network which can obtain a mapping from inputs to outputs by competitive learning and supervised learning. In the conventional counterpropagation, network outputs axe obtained by sum of outputs of units in previous layer, thus it is not effective to apply the counterpropagation to the system including heavy nonlinearity. In order to develop modeling ability, we employ neo fuzzy neurons, which are neuron models with nonlinear synapses, instead of sum for obtaining network outputs. The effectiveness and the validity of the proposed modified counterpropagation are verified by applying it to system modeling.