I am using a genetic algorithm to solve an optimization problem. However, I want a mating process that enables to create new solutions that are outside the range of current solutions. In the book: Practical Genetic Algorithms I have seen:
$$p_\mbox{new} = B*(p_1-p_2) + p_1$$
where $p_\mbox{new}$ is the new solution, and $p_1$ and $p_2$ are two different current solutions. $B$ is a random number between $[0,1]$.
But the problem is that this enables negative value, and my cromosomes can't be negative. I thought about the following:
$$p_\mbox{new} = A*p_1+B*p_2$$
where A and B are random numbers between [0,1]
Has this method been used before? Any opinion if it can work?
The other idea I had was just to used the one of Practical Genetic and take the absolute value, i.e.: $$ p_\mbox{new} = \left| B*(p_1-p_2) + p_1 \right| $$