While studying genetic algorithms, I've come across various crossover operators for the TSP problem. Some of these take two parents and produce 2 children (offspring), such as the alternating edge operator, for example. Others, such as partially mapped crossover (PMX) or sequential constructive crossover (SCX) only produce one child.
Now my question is: how does this relate to the crossover probability? More precisely: I'm adapting an implementation that uses a crossover operator that produces 2 children, and am replacing that operator with an operator that produces only 1 child (PMX). What should be altered in the implementation of the crossover step? Should I double the crossover probability, since only 1 new individual is produced now, every time crossover happens, instead of 2? I'm a bit lost here and cannot seem to find clear information on the difference between crossover operators that produce 1 child and other that produce 2.
Any help is much appreciated!
