[kepler-users] Automate workflows?

Thu Sep 19 08:28:13 PDT 2013

Hello all,

   I have been following the conversation, and since the subject opened up a bit I would also mention VisTrails.org.  They do not call it parameter sweep, but in actuality that is what they do as well.  Having said this VisTrails is out of the box geared toward visualizations (primarily because of the built-in modules), however it is very extensible.  I have looked at both Kepler and Vistrails.  I am a java programmer and had to step in to python as I chose Vistrails over Kepler.  I have no bias toward any tool and continue to follow the advances of Kepler.  Taverna was primarily web services oriented which did not fit our use case and target community.   Bottom line is that no workflow tool is a 100% fit. While I was reticent about learning the "plugin" aspects of any one tool due to time constraints, I found it the key to making it work.
   Having said this what Kepler has that Vistrails does not is the concept of Directors.  These, as best as I can tell, provide a variety of workflow "firing" models.  If you depend on that you probably need to stick with Kepler.

Eduardo

From: Christopher Brooks <cxh at eecs.berkeley.edu<mailto:cxh at eecs.berkeley.edu>>
Date: Thursday, September 19, 2013 10:12 AM
To: Muhammad Usama <11mscsmusama at seecs.edu.pk<mailto:11mscsmusama at seecs.edu.pk>>
Cc: "kepler-users at kepler-project.org<mailto:kepler-users at kepler-project.org>" <kepler-users at kepler-project.org<mailto:kepler-users at kepler-project.org>>
Subject: Re: [kepler-users] Automate workflows?

You mean there are workflow systems other than Kepler and Ptolemy?  :-)  I had no idea.

Actually, I don't know of other systems for parameter sweeps, but I'm sure that they are out there.  Simulink can probably do them.  You could look at Taverna.  The parameter sweeps page below lists https://lists.eecs.berkeley.edu/sympa/arc/ptolemy-hackers/2008-01/msg00007.html, which discusses STELLA.  See http://ptolemy.eecs.berkeley.edu/mailing-lists/split/0605/Re:_Berkeley_Madonna_and_System_Dynamics12 for a discussion about System Dynamics.

_Christopher

On 9/19/13 6:17 AM, Muhammad Usama wrote:
Thank you for reply. Parameter Sweeps is a new word for me. I didn't find this word while looking for a way to automate wokflows. But now I do and now I am reading about Parameter Sweeps.

Other than Kepler, do you know any other workflow system which can do parameter sweeps until the required results are found?

On Wed, Sep 18, 2013 at 10:18 PM, Christopher Brooks <cxh at eecs.berkeley.edu<mailto:cxh at eecs.berkeley.edu>> wrote:
Hi Muhammad,
This is an excellent question and something I've been meaning to try to answer in the Ptolemy II faq.

I just added something to the Ptolemy II faq about this:
http://ptolemy.eecs.berkeley.edu/ptolemyII/ptIIfaq.htm#parameterSweeps
says:
4.11 How do a do a parameter sweep?
A Parameter Sweep is when a model is run a number of times with different parameters to perhaps find an optimal solution. For details, see http://chess.eecs.berkeley.edu/ptexternal/wiki/Main/ParameterSweeps<http://chess.eecs.berkeley.edu/ptexternal/wiki/Main/ParameterSweeps#in_browser>.
http://chess.eecs.berkeley.edu/ptexternal/wiki/Main/ParameterSweeps
says:

Parameter Sweeps

A Parameter Sweep is when a model is run a number of times with different parameters to perhaps find an optimal solution.

These two articles in Wikipedia might be of interest.

  *   Design of Experiments (DoE)<https://en.wikipedia.org/wiki/Design_of_experiments> (Wikipedia)
  *   Monte Carlo Simulation<https://en.wikipedia.org/wiki/Monte_carlo_simulation> (Wikipedia)

One way to do optimization in Ptolemy II would be to do a parameter sweep on a model by using the higher order component (HOC) actors such as ModelReference.

See the ModelReference actor:

  *   Docs: http://sisyphus.eecs.berkeley.edu:8079/hudson/job/ptII/javadoc/ptolemy/actor/lib/hoc/ModelReference.html
  *   Demo: ptolemy/actor/lib/hoc/demo/ModelReference/ModelReference.xml

and RunCompositeActor

  *   http://sisyphus.eecs.berkeley.edu:8079/hudson/job/ptII/javadoc/ptolemy/actor/lib/hoc/RunCompositeActor.html

Tests:

  *   ptII/ptolemy/actor/lib/hoc/test/auto/DEInPNInRunCompositeActor.xml
  *   ptII/ptolemy/actor/lib/hoc/test/auto/RunCompositeActor.xml
  *   ptII/ptolemy/actor/lib/hoc/test/auto/RunCompositeActor2.xml
  *   ptII/ptolemy/actor/lib/hoc/test/auto/RunCompositeActor3.xml
  *   ptII/ptolemy/actor/lib/hoc/test/auto/RunCompositeActor4.xml

It is also possible to use a state machine:

  *   $PTII/ptolemy/domains/modal/demo/MultipleRuns/MultipleRuns.xml

Another way is to invoke the model multiple times from the command line and pass in different command line arguments. Properly escaping a string can be tricky.

  *   See http://ptolemy.eecs.berkeley.edu/ptolemyII/ptIIfaq.htm#invoking%20Ptolemy%20II?

Ptolemy/Kepler papers that cover Parameter Sweeps

  *   Jianwu Wang, "Distributed Execution Architectures in Ptolemy II<http://chess.eecs.berkeley.edu/pubs/806/>," 9th Biennial Ptolemy Miniconference, February 16, 2011
  *   Jianwu Wang, Altintas, I., Hosseini, P.R., Barseghian, D., Crawl, D., Berkley, C., Jones, M.B., "Accelerating Parameter Sweep Workflows by Utilizing Ad-hoc Network Computing Resources: an Ecological Example<http://users.sdsc.edu/%7Ejianwu/JianwuWang_files/Accelerating%20Parameter%20Sweep%20Workflows%20by%20Utilizing%20Ad-hoc%20Network%20Computing%20Resources%20-%20an%20Ecological%20Example%20%28SWF%202009%29.pdf>," Services - I, 2009 World Conference on, 6-10 July 2009. DOI<http://dx.doi.org/10.1109/SERVICES-I.2009.9>
  *   David Abramson, Blair Bethwaite, Colin Enticott, Slavisa Garic, Tom Peachey, "Parameter Space Exploration Using Scientific Workflows<http://link.springer.com/chapter/10.1007%2F978-3-642-01970-8_11>," Computational Science – ICCS 2009, Lecture Notes in Computer Science Volume 5544, 2009, pp 104-113.

Ptolemy/Kepler Mailing List mentions of Parameter Sweeps

  *   Monte Carlo Simulation In BCVTB<https://groups.google.com/forum/#%21msg/bcvtb/O0yofsqczOs/lNK1skgh5iQJ> October, 2012.
  *   Re: [kepler-users] System Dynamics Actor?<https://lists.eecs.berkeley.edu/sympa/arc/ptolemy-hackers/2008-01/msg00007.html> Jan. 2008.

On 9/18/13 6:14 AM, Muhammad Usama wrote:
Is there any way I can automate workflows? e.g. set the range of results which are required and then the workflow executes in loop and on each iterations some parameters are adjusted until the results meet the defined criteria?

Is there anyway I can do that in kepler? or is there any other system where I can do that?

--
Muhammad Usama
2011-NUST-MS-CS-009

_______________________________________________
Kepler-users mailing list
Kepler-users at kepler-project.org<mailto:Kepler-users at kepler-project.org>http://lists.nceas.ucsb.edu/kepler/mailman/listinfo/kepler-users

--
Christopher Brooks, PMP                       University of California
Academic Program Manager & Software Engineer  US Mail: 337 Cory Hall
CHESS/iCyPhy/Ptolemy/TerraSwarm               Berkeley, CA 94720-1774
cxh at eecs.berkeley.edu<mailto:cxh at eecs.berkeley.edu>, 707.332.0670<tel:707.332.0670>           (Office: 545Q Cory)

--
Muhammad Usama
2011-NUST-MS-CS-009

--
Christopher Brooks, PMP                       University of California
Academic Program Manager & Software Engineer  US Mail: 337 Cory Hall
CHESS/iCyPhy/Ptolemy/TerraSwarm               Berkeley, CA 94720-1774
cxh at eecs.berkeley.edu<mailto:cxh at eecs.berkeley.edu>, 707.332.0670           (Office: 545Q Cory)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nceas.ucsb.edu/kepler/pipermail/kepler-users/attachments/20130919/458edcda/attachment.html>