Genetic algorithms often stall in local optima, and restarting them is a common remedy. A new arXiv preprint proposes an adaptive restart rule that takes inspiration from ecology: the Schnabel Census, a method used to estimate animal population sizes. The rule is designed to decide when a genetic algorithm should be restarted, based on a census-like sampling logic.
According to the abstract, the Schnabel Census was originally developed for statistical estimation of animal population size, and the paper applies it to genetic algorithms. The authors describe the rule as adaptive, meaning it likely adjusts its behavior based on the algorithm's progress, though the abstract does not provide details on the implementation or experimental results.
Because the abstract is truncated, the full methodology and findings are not yet available from this source. The work is described as a new adaptive restart rule, but no comparative performance data is presented in the abstract.