Tracking Extrema in Dynamic Environments

摘要: Typical applications of evolutionary optimization involve the off-line approximation extrema static multi-modal functions. Methods which use a variety techniques to self-adapt mutation parameters have been shown be more successful than methods do not self-adaptation. For dynamic functions, interest is obtain but follow it as closely possible. This paper compares on-line tracking performance an program without self-adaptation against using self-adaptive Gaussian update rule over number dynamics applied simple function. The experiments demonstrate that for some effective while others detrimental.