MASON-INTEREST-L Archives

April 2004

MASON-INTEREST-L@LISTSERV.GMU.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Sean Luke <[log in to unmask]>
Reply To:
MASON Multiagent Simulation Toolkit <[log in to unmask]>
Date:
Tue, 6 Apr 2004 11:04:40 -0400
Content-Type:
text/plain
Parts/Attachments:
text/plain (31 lines)
Morten, MASON was -explicitly- designed to be used with ECJ; however I
don't believe we have actually done so yet.  Generally speaking, the
most straightforward approach is in your ECJ Problem subclass, you
create a MASON simulation (or use one you have hanging around
globally), replace its random number generator with
state.random[thread], and create one or more steppables which call your
Individual, then do something if the Individual returns something
they're supposed to do (maybe a vector or something). Then you just
pulse the MASON simulation until you think you're through, and analyze
the MASON results, convert to a fitness and return. Additionally, the
MASON simulation is placed in a certain location in the Problem so that
each of the functions and terminals can see it -- thus they're being
called within a steppable at this point and can just manipulate the
simulation as is appropriate.

Sean

On Apr 6, 2004, at 9:17 AM, Morten Revsbæk wrote:

> Hi
>
> I would like to take individuals from ECJ and turn them into agents in
> a
> mason simulation so that the fitness of each individual in ECJ is based
> upon how well they perform in the mason simulation. Does anyone know of
> any good way to combine ECJ and MASON in this way or maybe a pointer to
>  some litterature describing this?
>
> Regards
> --Morten Revsbæk

ATOM RSS1 RSS2