A fuzzy neural network tree with heuristic backpropagation learning
Yanqing Zhang, Fu-Lai Chung · 2003
To solve the curse of dimensionality of a conventional fuzzy neural network, a fuzzy neural network tree based on the normal fuzzy reasoning is proposed. The heuristic backpropagation learning algorithm using a divide-and-conquer method is developed to enhance learning quality in term of discovered knowledge, training error and prediction error. Simulations have shown that the fuzzy neural network tree is able to discover meaningful fuzzy rules with low training errors and low prediction errors. In the future, the fuzzy neural network tree will have more applications in large-scale data mining and data fusion, machine learning, and e-business.