Cerebellar model arithmetic computer with bacterial evolutionary algorithm and its hardware acceleration using FPGA
Masahiro Miwa, Tomonori Hashiyama, Takeshi Furuhashi, S. Okuma · 2003
Evolutionary algorithms (EAs) have been applied to various combinatorial optimization problems. The Bacterial Evolutionary Algorithm (BEA) is an optimization method that incorporates special mechanisms inspired by natural phenomena of microbial evolution. The BEA has been applied to fuzzy modeling of nonlinear systems for efficient discovery of appropriate rules and parameter tunings. Cerebellar Model Arithmetic Computer (CMAC) is a neural network that imitates the human cerebellum. One of the advantages of CMAC is its fast learning capability. It is, however, difficult to find the proper partition of input space of CMAC. The paper proposes an approach to the determination of input space division of CMAC using BEA. Hardware implementation of CMAC using BEA is also done to accelerate the computation time. The CMAC/BEA hardware is 108 times faster in computation than the software run on a conventional processor.