Useful diversity via multiploidy

A multiploid genetic algorithm (GA) incorporates several candidates for each gene within a single genotype, and uses some form of dominance mechanism (most simply, an encoded choice) to decide which choice of each gene is active in the phenotype. We explore a simple multiploid model. Investigation with two simplified test problems is reported, respectively suggesting certain strengths and weaknesses of employing multiploidy. In particular, multiploidy appears useful in cases where attractive suboptima are profoundly Hamming distant from the true optimum, thus requiring a GA to recover substantial lost material in order to recover from suboptima. This is distinct from cases where a GA's difficulty in solving a problem is, for example, more concerned with appropriately combining genetic material than finding it.