A COMBINE ALGORITHM FOR A CMAC NETWORK
Selahattin Sayıl · Pamukkale University Journal of Engineering Sciences · 2011
The performance of a CMAC neural network depends on the training algorithms and the selection of inputpoints. Papers have been published that explain CMAC algorithms but little work has been done to improveexisting algorithms. In this paper, the existing algorithms are first explained and then compared usingcomputational results and the algorithm properties. Improvements are made to the recommended MaximumError by using a Combine Algorithm approach. In this method, CMAC network is first trained byusing Neighborhood Training and then trained by Maximum Error for fine-tuning ofCMAC network. Faster initial convergence is achieved for the recommended Maximum Error Algorithm. Thisapproach may reduce the training time and accelerate the initial learning which is very important in manycontrol applications.