ECJ-INTEREST-L Archives

April 2005

ECJ-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:
Harri Salakoski <[log in to unmask]>
Reply To:
ECJ Evolutionary Computation Toolkit <[log in to unmask]>
Date:
Tue, 12 Apr 2005 18:17:27 +0300
Content-Type:
text/plain
Parts/Attachments:
text/plain (124 lines)
> Also I would like to
> execute the runs sequentially so that I could terminate if a good enough
> individual is found.  My problem is I don't see an obvious place to put an
> outer loop in ECJ without changing quite a few things. Does anyone know a
> simple solution?

I have coded ecj in separate branch now several years (XML things and such,
latest item is jdk1.5 porting). Quite lately I added separate goal package
which tries to figure out are research goals filled. If not it starts new
simulations by variying parameters (depends used goal strategy), it kind of
tries automate my research. Or kind of that hypotetical future I can focus
to inspect results, just define varied variables and generated graphs.
Another thing for it is that it should optimise used resources so that it
uses all computing power available to make results better or simulations
bigger if there is time. It kind of works, have been happy for this goal
package and keep it. Something similar "metasim", "thread manager", "sim
manager" or "goal" package could fit for ecj also to help scientist and I
think there is no such item yet.

t. Harri


----- Original Message -----
From: "Hirsch Laurence" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Tuesday, April 12, 2005 3:57 PM
Subject: Re: Multiple Runs


> Returning to the discussion below after a long break from ECJ.  Both the
> suggestions work fine, however ideally I would like to implement multiple
> runs in Java (i.e. operating system independent).  Also I would like to
> execute the runs sequentially so that I could terminate if a good enough
> individual is found.  My problem is I don't see an obvious place to put an
> outer loop in ECJ without changing quite a few things. Does anyone know a
> simple solution?
>
> Thanks,
>
>
> Laurie
>
>
> -----Original Message-----
> From: Sean Luke [mailto:[log in to unmask]]
> Sent: 09 July 2004 17:15
> To: [log in to unmask]
> Subject: Re: Multiple Runs
>
> They'd be equal more or less, but of course they'd all write out to one
> big stat file by default.  I've not done it with 50 subpopulations
> before -- I always tended to reserve subpopulations for other purposes.
>  But I think it'd probably work just fine.
>
> Sean
>
> On Jul 9, 2004, at 11:59 AM, T. S. wrote:
>
>> Out of curiosity how is this better than running say 50 sub-populations
>> where there is no inter sub-population breeding? Or would they be
>> equal,
>> except for the need of 50 seeds?
>>
>> -Travis
>>
>>
>>> From: Sean Luke <[log in to unmask]>
>>> Reply-To: ECJ Evolutionary Computation Toolkit
>>> <[log in to unmask]>
>>> To: [log in to unmask]
>>> Subject: Re: Multiple Runs
>>> Date: Fri, 9 Jul 2004 11:42:14 -0400
>>>
>>> This little shell script will run ECJ in single-threaded mode 50 times
>>> using the seeds START+1 through START+50, and the file MYFILE.PARAMS.
>>> Each stat file will be outputted as 1.out.stat through 50.out.stat in
>>> your current directory.  You can pass in additional parameters using
>>> the typical -p foo=bar format.  Save it as "ecj" or something and put
>>> it in an appropriate place.
>>>
>>> USAGE
>>>        ecj MYFILE.PARAMS START [-p param]*
>>>
>>> EXAMPLE.  Will run ant 50 times with just 4-generation runs,
>>>             starting at seed = 101.
>>>        cd ec/app/ant
>>>        ecj ant.params 100 -p generation=4
>>>
>>> THE SCRIPT
>>> -------------------------------
>>> #! /bin/tcsh
>>> @ z=$2
>>> foreach r (1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
>>> 24 25 26 27 28 29 30 31 32 33 34 45 36 37 38 39 40 41 42 43 44 45 46
>>> 46
>>> 47 48 49 50)
>>>        @ seed=$z + $r
>>>        java ec.Evolve -file $1 \
>>>                       -p seed.0=$seed \
>>>                       -p stat.file=\$${r}.out.stat \
>>>                                 $argv[*]
>>>        end
>>> echo All Done!
>>> -------------------------------
>>>
>>>
>>> Sean
>>>
>>>
>>>
>>>
>>> On Jul 9, 2004, at 9:03 AM, Hirsch Laurence wrote:
>>>
>>>> Is there a recommended way to implement multiple runs in ECJ?
>>>>
>>>> Thanks,
>>>>
>>>> Laurie
>>
>> _________________________________________________________________
>> MSN 9 Dial-up Internet Access helps fight spam and pop-ups - now 2
>> months
>> FREE! http://join.msn.click-url.com/go/onm00200361ave/direct/01/

ATOM RSS1 RSS2