Blind Equalizer Design Using Genetic Programming to Evolve the Cost Function

Along with the advent of wireless mobile communications, blind equlizers are becoming a key component for systems that demand reliability and robustness. Central to the design of blind equalizers is the cost function that the adaptive lter algorithm minimizes. Mean-square errors are found no longer applicable as information-bearing source sequences are not available to blind equalizers. We believe that it becomes heuristic in nature to seek a cost function that is simple to implement in hardware while eeective in establishing convergence. Genetic programming is used to explore the function space systematically and breed the cost functions that can eventually meet operational speciications. The heuristic functions we have obtained outperform the existing results to date such as CMA, WGI, Sato, and Godard, by at least an order of magnitude in convergence time. Unlike CMA, our algorithm also achieves universal convergence independent of initialization.