基于自适应学习速率法的补偿模糊神经网络

为克服常规网络收敛速度慢、无法结合专家知识等缺点,引入补偿模糊神经元,结合模糊系统强大的知识表达能力和神经网络优秀的自学习能力,并利用自适应学习速率法动态地改变学习率.提出了一种新型的基于自适应学习速率法的补偿模糊神经网络,并将其应用到实际例子中.结果证明,它不仅能在线适当调整参数,还能动态地优化相应的模糊推理,加快训练速度.

Design and Application of Compensatory Fuzzy Neural Network Based on Adaptive Learning Rate Method

To overcome the disadvantages of general networks such as slow convergence speed and being unable to combine with the expert knowledge etc,the authors introduces compensatory fuzzy neural cells,integrate the powerful knowledge expressiveness of fuzzy system and the excellent self-learning of neural network,and present a novel Compensatory Fuzzy Neural Network(CFNN) based on Adaptive Learning Rate Method which changes the learning rate using Adaptive Learning Rate Method in dynamic way.Finally this method is applied to the actual case.The result proves that it not(only) can adjust parameters properly on line,but also can optimize relevant fuzzy reasoning in dynamic way,fasten training rate.