Genetic algorithms for maximum likelihood parameter estimation

The authors introduce genetic algorithms (GAs) to solve the exact maximum-likelihood equations arising in a typical signal processing problem. GAs are an efficient and highly parallel way of maximizing complex, multimodal, multivariable functions. They are based on concepts borrowed from natural genetic evolution, involving reproduction, crossover, and mutation of a continuously evolving population of parameter estimates. An outline is presented of a GA for maximizing a Gaussian likelihood function, and it is shown that it can outperform existing high-performance parameter estimation algorithms under difficult conditions