id,summary,reporter,owner,description,type,status,priority,milestone,component,version,resolution,keywords,cc 1682,Probabilistic functional crossover,bburlacu,gkronber,"This crossover operator increases the probability of a beneficial recombination event by (probabilistically) weighting the nodes in the genome based on their ''behavioral'' rather than ''structural'' similarity. The main idea is that when a node ''i'' is randomly chosen from the first parent, the second parent is scanned to find the node ''j'' that has the closest range to that of the chosen node in the first parent, based on a sort of ''behavioral distance'' that takes into account the minimum and maximum values computed by the two nodes during computation. However, instead of being greedy (the case for the deterministic functional crossover), the procedure (which involves computing all distances between node ''i'' and every node ''k'' from the second parent) assigns a probability for each node in the second parent to be selected as the second cross point, based on the inverted normalized value of the behavioral distance (so the lower the distance, the higher the probability). J.C. Bongard in [''1''] claims that this method outperforms genetic programming without crossover, random crossover, and a deterministic form of the crossover operator in the symbolic regression domain. [''1''] [http://dl.acm.org/citation.cfm?id=1830649 A Probabilistic Functional Crossover Operator for Genetic Programming], Proceeding, GECCO '10 Proceedings of the 12th annual conference on Genetic and Evolutionary Computation",feature request,closed,medium,HeuristicLab 3.3.7,Problems.DataAnalysis.Symbolic,3.3.7,done,,mkommend