A.12 SOFTWARE-EVOLUTION-LIBRARY/COMPONENTS/MULTI-OBJECTIVE

Variable: *pareto-comparison-set-size* [sel/cp/multi-objective]

Size of comparison set for Pareto tournaments.

Function: crowding-distance [sel/cp/multi-objective] software

Compute crowding distance for software by comparing with whole population.

Function: dominates-all [sel/cp/multi-objective] predicates comparison-set candidate

Does candidate dominate all of comparison-set?

predicates is a list of functions, of equal length to the fitness values, which are used to compare each element of the fitness values.

Function: multi-objective-scalar [sel/cp/multi-objective] fitness

docfixme

Function: pareto-selector [sel/cp/multi-objective] candidates &key predicate comparison-set

Return members of candidates which dominate a random comparison-set.

Set ‘*tournament-selector*’ to ‘pareto-selector’ to use Pareto selection in tournaments.

Function: pick-least-crowded [sel/cp/multi-objective] candidates &key predicate

Pick candidate with the greatest crowding distance.

Crowding distance is a fitness sharing metric adapted from nsga-ii. For each fitness component, it sums the distance between an individual and its nearest neighbors. Individuals with the greater crowding distance are in sparse areas of the fitness landscape and should be preferred.