Kernel CMAC model with fast convergence

In order to improve the speed and accuracy of the on-line learning neural network, a credit assignment concept based on kernel CMAC was proposed, where the credit was supposed to be the activated hypercubes previous learning times in the kernel space. The correcting amounts of errors are directly proportional to the inversion of the learned times of the addressed hypercubes. With this idea, the CMAC training regulation was designed in kernel space to improve the network learning speed and accuracy, and improve the modeling capability. The simulation result shows that the improved CMAC neural network in kernel space is faster and more accurate than the conventional CMAC.