Genetic Algorithm and Non-Linear Constraint Solver in Series and Parallel

Hi,

I am conducting an optimisation with bound, mixed integer and non-linear constraints with the genetic algorithm solver. My objective and constraint function are costly to evaluate and share computation. I have successfully placed the objective and constraint constraints in the same function as per https://www.mathworks.com/help/optim/ug/objective-and-nonlinear-constraints-in-the-same-function.html.

However I would like to understand how effective this approach is in each of the following cases. I have included an example answers of the type of thing I hope to learn for the serial cases (not necessarily correct!). However, I cannot do the same for the parallel case as I don't understand how the computation is split up.

• Serial computation + Augmented Langrangian Algorithm. Example answer: constraints and fitness are evaluated serially for a given individual therefore storing the last individual effectively reduces the runtime.
• Serial computation + Penalty/Integer Algorithm. Example answer: all constraints are evaluated and then fitness for individuals that satisfy constraints are evaluated therefore storing the last individual is not an effective way of reducing the runtime.
• Parallel computation + Augmented Langrangian Algorithm
• Parallel computation + Penalty/Integer Algorithm

I'd hope to learn how the problem is split up in parallel and whether there a way to effectively share information between the cosntraint and fitness function in all of the above cases.

Best,

Ryan