Thank you for these advices!

Noémie



--- In openbeagle@yahoogroups.com, Christian Gagné <cgagne@...> wrote:
>
> Hi,
>
> The thing here is to increase the minimum depth of some parameters, as
> the spambase example typing require that the trees are at least of 3
> levels (or something similar, I don't remember precisely). Make the
> exercise, looking at the typing, it is not possible to have trees with
> less that number of levels.
>
> My intuition is that you should increase the value of
> "gp.init.mindepth" to something like 3 or 4, in order to get GP trees
> with enough levels to allow different crossover and mutation operators
> to apply. That's something I did in the past. Otherwise, I would give
> you usual advice to follow, that is to use a debugger and take the
> time to observer where the thing get stuck. It can be quite
> instructive, really, consider doing it.
>
> Christian
>
>
> On Jul 14, 2009, at 4:47 PM, noemie.perona wrote:
>
> > Thank you for your quick answer.
> > The thing is I use STGP in my project (and I can't make it work)
> > that's why I've tested Spambase example .
> > I looked at sp-evolver again but everything seems fine to me but
> > maybe I misunderstood something.
> >
> > Here is sp-evolver :
> > <?xml version="1.0" encoding="ISO-8859-1"?>
> > <Beagle version="2.2.0">
> > <Evolver>
> > <BootStrapSet>
> > <DB-BootStrapBeginOp/>
> > <IfThenElseOp parameter="db.restart" value="1">
> > <PositiveOpSet/>
> > <NegativeOpSet>
> > <GP-InitHalfConstrainedOp/>
> > <DB-DistribEvalOp/>
> > </NegativeOpSet>
> > </IfThenElseOp>
> > <DB-BootStrapEndOp/>
> > </BootStrapSet>
> > <MainLoopSet>
> > <DB-MainLoopBeginOp/>
> > <SelectTournamentOp repropb="ec.repro.prob"/>
> > <GP-CrossoverConstrainedOp matingpb="gp.cx.indpb"
> > distrpb="gp.cx.distrpb"/>
> > <GP-MutationStandardConstrainedOp mutationpb="gp.mutstd.indpb"
> > maxdepth="gp.mutstd.maxdepth"/>
> > <GP-MutationShrinkConstrainedOp mutationpb="gp.mutshrink.indpb"/>
> > <GP-MutationSwapConstrainedOp mutationpb="gp.mutswap.indpb"
> > distrpb="gp.mutswap.distrpb"/>
> > <GP-MutationSwapSubtreeConstrainedOp
> > mutationpb="gp.mutsst.indpb" distrpb="gp.mutsst.distrpb"/>
> > <DB-DistribEvalOp/>
> > <MigrationRandomRingOp/>
> > <GP-StatsCalcFitnessSimpleOp/>
> > <TermMaxGenOp/>
> > <DB-MilestoneWriteOp/>
> > <DB-MainLoopEndOp/>
> > </MainLoopSet>
> > </Evolver>
> > <Register>
> > <Entry key="db.app.name">dbeagle-app</Entry>
> > <Entry key="db.comm.retry">5</Entry>
> > <Entry key="db.cruncher.dumpconf"/>
> > <Entry key="db.evolver.compression">0</Entry>
> > <Entry key="db.restart">0</Entry>
> > <Entry key="db.server.ip">192.168.1.2</Entry>
> > <Entry key="db.server.port">9123</Entry>
> > <Entry key="ec.conf.file"/>
> > <Entry key="ec.hof.demesize">0</Entry>
> > <Entry key="ec.hof.vivasize">1</Entry>
> > <Entry key="ec.init.seedsfile"/>
> > <Entry key="ec.mig.interval">1</Entry>
> > <Entry key="ec.mig.size">5</Entry>
> > <Entry key="ec.pop.size">4</Entry>
> > <Entry key="ec.rand.seed">2</Entry>
> > <Entry key="ec.rand.state">0</Entry>
> > <Entry key="ec.repro.prob">0.1</Entry>
> > <Entry key="ec.sel.tournsize">2</Entry>
> > <Entry key="ec.term.maxgen">2</Entry>
> > <Entry key="gp.cx.distrpb">0.9</Entry>
> > <Entry key="gp.cx.indpb">0.9</Entry>
> > <Entry key="gp.init.maxdepth">5</Entry>
> > <Entry key="gp.init.mindepth">2</Entry>
> > <Entry key="gp.mutshrink.indpb">0.05</Entry>
> > <Entry key="gp.mutsst.distrpb">0.5</Entry>
> > <Entry key="gp.mutsst.indpb">0</Entry>
> > <Entry key="gp.mutstd.indpb">0.05</Entry>
> > <Entry key="gp.mutstd.maxdepth">5</Entry>
> > <Entry key="gp.mutswap.distrpb">0.5</Entry>
> > <Entry key="gp.mutswap.indpb">0.05</Entry>
> > <Entry key="gp.tree.maxdepth">17</Entry>
> > <Entry key="gp.try">2</Entry>
> > <Entry key="lg.console.enabled">1</Entry>
> > <Entry key="lg.file.name">beagle.log</Entry>
> > <Entry key="lg.log.level">3</Entry>
> > <Entry key="lg.show.class">0</Entry>
> > <Entry key="lg.show.level">0</Entry>
> > <Entry key="lg.show.type">0</Entry>
> > <Entry key="ms.write.interval">1</Entry>
> > <Entry key="ms.write.over">0</Entry>
> > <Entry key="ms.write.perdeme">0</Entry>
> > <Entry key="ms.write.prefix">beagle</Entry>
> > </Register>
> > </Beagle>
> >
> >
> > Noémie
> >
> > --- In openbeagle@yahoogroups.com, Christian Gagné <cgagne@> wrote:
> >>
> >> Hi Noemie,
> >>
> >> By experience, it is common to get spambase example to stall. The
> >> problem is that this is an example of strongly-typed GP (STGP), and
> >> that with a bad configuration, the typing can be never valid under
> >> some circumstances. The STGP is implemented in version 2.2.0 with
> >> brute-force, when some configuration is not feasible according to the
> >> typing constraints, the program is sometime not able to figure it. In
> >> these circumstances, it just try again and again, indefinitely.
> >>
> >> If you don't want to do STGP, I would advice you to avoid it, given
> >> these considerations. If you insist, I would advice you to look at
> >> the
> >> sp-evolver.conf file, which is probably using parameters that are not
> >> correct.
> >>
> >> Best,
> >>
> >> Christian
> >>
> >>
> >>
> >>
> >> On Jul 14, 2009, at 4:25 PM, noemie.perona wrote:
> >>
> >>> Hello,
> >>>
> >>> I tested Spambase of Dbeagle but I only got the first population.
> >>> What can be the problem? Do you think it is related to the evolver?
> >>>
> >>> Here is the result produced by the evolver:
> >>>
> >>> <?xml version="1.0" encoding="ISO-8859-1"?>
> >>> <Beagle version="2.2.0">
> >>> <Logger>
> >>> <Log>Open BEAGLE, version 2.2.0</Log>
> >>> <Log>Logger initialized, using log level 3 (general
> >>> informations)</Log>
> >>> <Log>Reading evolver from file named "sp-evolver.conf"</Log>
> >>> <Log>Parsing command-line "/global/home/openbeagle/dbeagle/
> >>> examples/GP/spambase/spambase/spambase-client -OBec.conf.file=sp-
> >>> evolver.conf"</Log>
> >>> <Log>Reading parameter file named "sp-evolver.conf"</Log>
> >>> <Log>Starting an evolution</Log>
> >>> <Log>Current date and time: 12:54:35 14 Jul 2009</Log>
> >>> <Log>Applying bootstrap operators to the 1st deme</Log>
> >>> <Log>Applying main-loop operators to the 1st deme</Log>
> >>> <Log>
> >>> <Stats id="deme1" generation="0" popsize="4">
> >>> <Item key="processed">4</Item>
> >>> <Item key="total-processed">4</Item>
> >>> <Measure id="fitness">
> >>> <Avg>0.533125</Avg>
> >>> <Std>0.0790932</Std>
> >>> <Max>0.6025</Max>
> >>> <Min>0.42</Min>
> >>> </Measure>
> >>> <Measure id="treedepth">
> >>> <Avg>3.75</Avg>
> >>> <Std>1.5</Std>
> >>> <Max>5</Max>
> >>> <Min>2</Min>
> >>> </Measure>
> >>> <Measure id="treesize">
> >>> <Avg>8</Avg>
> >>> <Std>5.94418</Std>
> >>> <Max>16</Max>
> >>> <Min>3</Min>
> >>> </Measure>
> >>> </Stats>
> >>> </Log>
> >>> <Log>
> >>> <Stats id="vivarium" generation="0" popsize="4">
> >>> ...
> >>> </Stats>
> >>> </Log>
> >>> <Log>Writing milestone file "beagle_g0.obm.gz"</Log>
> >>> <Log>Applying main-loop operators to the 1st deme</Log>
> >>>
> >>>
> >>> Noémie
> >>>
> >>>
> >>>
> >>>
> >>> --- In openbeagle@yahoogroups.com, Christian Gagné <cgagne@> wrote:
> >>>>
> >>>> Hi Noémie,
> >>>>
> >>>> What is important point is that when you launch a binary on a given
> >>>> computer it should have access to the configuration file, either at
> >>>> default location (current directory) or path given on the command-
> >>>> line.
> >>>>
> >>>> It doesn't really matter if the configuration file is the same
> >>>> physical file for each node/client. You can share the configuration
> >>>> file through a distributed filesystem (e.g. NFS on Unix), or make a
> >>>> different copy on the local filesystem of each node. What is
> >>>> important
> >>>> is to launch the client with the correct reference to the
> >>>> configuration file to use.
> >>>>
> >>>> Christian
> >>>>
> >>>>
> >>>>
> >>>>
> >>>> On Jun 26, 2009, at 1:44 PM, noemie.perona wrote:
> >>>>
> >>>>> Thank you Christian,
> >>>>>
> >>>>> I am not sure about something :
> >>>>> Where should be placed configuration files of clients? They can be
> >>>>> at the same place (for example in the current directory), or each
> >>>>> configuration file has to be on the same node as the client?
> >>>>>
> >>>>> Noémie
> >>>>>
> >>>>> --- In openbeagle@yahoogroups.com, Christian Gagné <cgagne@>
> >>>>> wrote:
> >>>>>>
> >>>>>> Hi Noémie,
> >>>>>>
> >>>>>> On Jun 25, 2009, at 1:42 PM, noemie.perona wrote:
> >>>>>>
> >>>>>>> Hello everybody,
> >>>>>>>
> >>>>>>> I am trying to use Distributed Beagle but I have some troubles
> >>>>>>> with
> >>>>>>> it.
> >>>>>>> If you use a script to start the server, the evolver and then
> >>>>>>> crunchers on several nodes of the cluster,how do you stop them
> >>>>>>> when
> >>>>>>> the evolver is done? by hand ( with Ctrl-c)? with your script?
> >>>>>>> Is
> >>>>>>> there a way for them to terminate themself?
> >>>>>>
> >>>>>> Ctrl-c is one of the good way to stop the server. The server is
> >>>>>> coded
> >>>>>> to support elegant termination through Ctrl-c. Another approach
> >>>>>> is to
> >>>>>> do a 'kill' on the process (see good unix manual for more details
> >>>>>> on
> >>>>>> that).
> >>>>>>
> >>>>>> Christian
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>> Noémie
> >>>>>>>
> >>>>>>>
> >>>>>>> --- In openbeagle@yahoogroups.com, Christian Gagné <cgagne@>
> >>>>>>> wrote:
> >>>>>>>>
> >>>>>>>> Hi Stig,
> >>>>>>>>
> >>>>>>>> Changes are quite important, with a reorganisation of the
> >>>>>>>> source
> >>>>>>>> code
> >>>>>>>> in smaller package, and a new build system. Until the next
> >>>>>>>> release, I
> >>>>>>>> freeze any new features to focus on completing and testing the
> >>>>>>>> new
> >>>>>>>> changes. Multithreaded fitness evaluation is not planned to be
> >>>>>>>> part
> >>>>>>>> of
> >>>>>>>> that release, as it requires a great deal of modification. It
> >>>>>>>> will
> >>>>>>>> probably be better of the following release (the one after the
> >>>>>>>> next
> >>>>>>>> one).
> >>>>>>>>
> >>>>>>>> Cheers,
> >>>>>>>>
> >>>>>>>> Christian
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> On May 18, 2009, at 4:32 AM, <stig.petersen@> <stig.petersen@
> >>>>>>>>> wrote:
> >>>>>>>>
> >>>>>>>>> Hi Christian,
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> I am very excited about the new framework version, which you
> >>>>>>>>> plan to
> >>>>>>>>> release in the summer :]
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Is there a list of planned changes somewhere? In particular,
> >>>>>>>>> will it
> >>>>>>>>> include changes to how you deal with multithreading to make it
> >>>>>>>>> easier to fully utilize multi-core CPUs?
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Thanks
> >>>>>>>>>
> >>>>>>>>> -Stig
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> From: openbeagle@yahoogroups.com [mailto:openbeagle@yahoogroups.com
> >>>>>>>>> ]
> >>>>>>>>> On Behalf Of Christian Gagné
> >>>>>>>>> Sent: 18 May 2009 02:57
> >>>>>>>>> To: openbeagle@yahoogroups.com; noemie.perona
> >>>>>>>>> Subject: Re: [OpenBEAGLE] use of Distributed Beagle
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Hi Noemie,
> >>>>>>>>>
> >>>>>>>>> The development version of Open BEAGLE is not documented a
> >>>>>>>>> lot,
> >>>>>>>>> although many element of previous documentation still apply.
> >>>>>>>>> You
> >>>>>>>>> can
> >>>>>>>>> take a look at it. But note that the version that have been
> >>>>>>>>> released
> >>>>>>>>> is limited. I am currently resuming working on it, adding new
> >>>>>>>>> stuff,
> >>>>>>>>> including distribution code.
> >>>>>>>>>
> >>>>>>>>> We are looking to release a new version of the framework
> >>>>>>>>> toward
> >>>>>>>>> mid/
> >>>>>>>>> end of this summer. That should include distribution code.
> >>>>>>>>>
> >>>>>>>>> Of course, there is nothing that stop you from installing a
> >>>>>>>>> previous
> >>>>>>>>> version of Open BEAGLE that is compatible with Distributed
> >>>>>>>>> BEAGLE.
> >>>>>>>>> All
> >>>>>>>>> versions are available on sourceforge.
> >>>>>>>>>
> >>>>>>>>> Christian
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> ____________________________________________________________________
> >>>>>>>>> Check out NEoN, a ground-breaking digital arts festival taking
> >>>>>>>>> place
> >>>>>>>>> in Dundee on 13th-14th November 2009.
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> DISCLAIMER
> >>>>>>>>>
> >>>>>>>>> This message and any attachments contain privileged and
> >>>>>>>>> confidential
> >>>>>>>>> information intended for the use of the addressee named above.
> >>>>>>>>> If
> >>>>>>>>> you are not the intended recipient of this message, you are
> >>>>>>>>> hereby
> >>>>>>>>> notified that any use, dissemination, distribution or
> >>>>>>>>> reproduction
> >>>>>>>>> of this message is prohibited. Please note that we cannot
> >>>>>>>>> guarantee
> >>>>>>>>> that this message or any attachment is virus free or that it
> >>>>>>>>> has
> >>>>>>>>> not
> >>>>>>>>> been intercepted and amended. The views of the author may not
> >>>>>>>>> necessarily reflect those of Realtime Worlds Ltd.
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> Realtime Worlds Ltd is registered in Scotland, number 225628.
> >>>>>>>>> Registered Office: 152 West Marketgait, Dundee, DD1 1NJ.
> >>>>>>>>>
> >>>>>>>>> [Non-text portions of this message have been removed]
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> ------------------------------------
> >>>>>>>>>
> >>>>>>>>> Yahoo! Groups Links
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>
> >>>>>>>> Christian Gagné
> >>>>>>>> http://vision.gel.ulaval.ca/~cgagne
> >>>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>> ------------------------------------
> >>>>>>>
> >>>>>>> Yahoo! Groups Links
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>> Christian Gagné
> >>>>>> http://vision.gel.ulaval.ca/~cgagne
> >>>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>> ------------------------------------
> >>>>>
> >>>>> Yahoo! Groups Links
> >>>>>
> >>>>>
> >>>>>
> >>>>
> >>>> Christian Gagné
> >>>> http://vision.gel.ulaval.ca/~cgagne
> >>>>
> >>>
> >>>
> >>>
> >>>
> >>> ------------------------------------
> >>>
> >>> Yahoo! Groups Links
> >>>
> >>>
> >>>
> >>
> >> Christian Gagné
> >> http://vision.gel.ulaval.ca/~cgagne
> >>
> >
> >
> >
> >
> > ------------------------------------
> >
> > Yahoo! Groups Links
> >
> >
> >
>
> Christian Gagné
> http://vision.gel.ulaval.ca/~cgagne
>