Evolutionary Algorithms with Competing Heuristics in Computational Statistics

The paper presents a new class of evolutionary algorithms based on the competition of different heuristics. The algorithm was applied to solving some optimization problems of computational statistics, namely to estimating the parameters of non-linear regression models, constrained M-estimates and optimizing the smoothing constants in the Winters exponential smoothing. The results showed that the evolutionary algorithm with competing heuristics can be successfully used in solving some global optimization problems of computational statistics.