Liviu Panait wrote:
> I have previously used ECJ and MASON together, and it is quite
> straightforward. In my approach, I created a new class that extended
> GPProblem (I wanted to used GP), and it contained a MASON simulation as
> a member. A new instance of the simulation was created at each
> evaluation. One of the parameters for constructing the simulation was
> the GPIndividual. This parameter was then passed forward to the agents
> in the simulation. Therefore, at each time step (or whenever the
> agents in the simulation were invoked to do their stuff), they had a
> pointer to the GPIndividual that needed to be evaluated, and took their
> decisions based on this individual. During the simulation, I was
> collecting various data used to assess the fitness.
> Unfortunately, the code used some preliminary version of the MASON
> simulator, and needs small tweaks to compile with the new version.
> Plus, it is poorly commented. However, I could send it to you as an
Thanks for your reply. It sounds pretty simple, but it could not harm to
have a look at your code before starting on my own. I would appreciate
it, if you mailed it to me.