I have also been developing on a seperate branch for a while now - what
you're doing sounds interesting - is it available publically anywhere?
Harri Salakoski wrote:
>> 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/
>>
>
--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.308 / Virus Database: 266.9.6 - Release Date: 11/04/2005
|
