Context-aware crossover

[bburlacu]

Given a randomly selected subtree from one parent, context-aware crossover will always find the best location to place the subtree in the other parent.

In their paper ([1]), Majeed and Ryan suggest that using standard crossover in the initial part of the run and then switching to context-aware crossover yields the best performance. They show that, across a range of standard GP benchmark problems, context-aware crossover produces a higher best fitness as well as a higher mean fitness, and even manages to solve the 11-bit multiplexer problem without ADFs. At the same time, the individuals produced this way are much smaller than standard GP, and far fewer individual evaluations are required, so GP achieves a higher fitness by evaluating fewer and smaller individuals.

The authors hypothesize that the order in which crossover operators are applied (for instance the standard and the context-aware ones) and the time of their application can have a big impact on the experiments outcome.

[1] ​Using Context-aware Crossover to Improve the Performance of GP, Proceeding, GECCO '06, Proceedings of the 8th annual conference on Genetic and evolutionary computation, ACM New York, NY, USA ©2006.

[bburlacu]

r7109: Implemented the context-aware crossover.

[bburlacu]

r7110: Submitted another variant of the context-aware crossover, called the DeterministicBestCrossover. This one selects a random crossover point and finds the best node from the other parent to be inserted there.

[mkommend]

Reviewing comments:

• selection of cutpoints
• reorder criteria for cutpoint selection
• swap method
• do not set the operator name in the default ctor as this is done by the ItemAttribute automatically
• improve descriptions

[bburlacu]

r7495: Simplified selection of cutpoints and reordered criteria for cutpoint selection