Hi Sunil, All,

Vedrin Jeliazkov wrote:
> Hi,
>
> I'm forwarding Sunil's mail to the list -- I hope somebody out there
> would provide some good answers to Sunil's and Surajit's queries.
>
> Sunil, Surajit, are you members of the OTDev list already?
>
> Micha, Barry, could you please make sure that Sunil and Surajit have
> access to the OTDev list and its archives (if possible)?
>
> Kind regards,
> Vedrin
>
> ---------- Forwarded message ----------
> From: sunil chawla <sunil at seascapelearning.com>
> Date: 2009/12/22
> Subject: Re: [OTDev] SW Development process...SL update
> To: Vedrin Jeliazkov <vedrin.jeliazkov at gmail.com>
> Cc: Barry Hardy <barry.hardy at douglasconnect.com>, surajit ray
> <mr.surajit.ray at gmail.com>
>
>
> Hi Vedrin
>
> Thanks for your thorough and quick responses. Following is a quick
> update from our side...
>
> Our team had been occupied predominantly on developing the MaxTox
> prediction algorithm. And as the API has started emerging recently,
> Surajit is working to conform to it.  A 1/1 discussion/Q&A point of
> contact would be most helpful to make quick progress. Should Surajit
> be talking with you?
>
>   
I would suggest sending queries to the development/partners list, thus
you usually will get more information back than from a single person.
And most, probably, other developers could also benefit from a
discussion on the mailing list.
> A status update  is as follows:
> Software to make Dictionaries ===> ready
>   

> Software to make models ===> R (random forests classification)
> Software to make predictions ====> ready
> Online version of Prediction ====> Work in progress (halfway done in JSP)
> Online version of Prediction using OpenTox API ====> Just Started
> Issues we are wrestling with:
> 1) Rapid evolution of API and quick turns on testing is presenting a
> moving target.
>   
This is true for all of us.  A more active participation from your side
in the API discussions will be helpful, otherwise we might just fail to
include in the API some essential for your software functionality.
> 2) Unclear about whether we should fit our algorithm into Blue Obelix
>   
It might be better suited to describe your algorithms in Algorithm types
ontology, suggested by TUM , which is expected to include machine
learning algorithms.  AFAIK Blue Obelisk ontology currently doesn't
include machine learning algorithms .

An initial version of algorithm types ontology is available here
http://opentox.org/data/documents/development/RDF%20files/AlgorithmTypes/view
.  Could you please have a look and then we will discuss where does your
algorithm fit?
> 3) Should we host our software on our own servers or yours? What is
> the recommended strategy?
>   
It's up to you.  So far every team hosts the software on its own
servers, which facilitates the development and  quick deployment.  In
your case, it seems the network connectivity to your server is not very
good currently  (see Vedrin's test results at
http://www.opentox.org/dev/testing/testresults ) and it might make sense
to deploy a stable version of the software somewhere else.  I will leave
the exact answers where and how to the OpenTox webmaster.
> 4) There is no prior experience with RDF/OWL, we will need some help here.
>   
All the partners started without prior experience with RDF/OWL.    I
have tried to put together some introduction and Java examples at
http://www.opentox.org/data/documents/development/RDF%20files/?searchterm=RDF%20files 
- did you look at those?

Could you provide more details on your software development process
(language, development platforms, specific problems encountered)?
> 5) Unable to make the R part of the code native to java as yet. Still
> trying a few things. This hinders our ability to provide the API
> interface.
>   
I am aware of at least one Java - R integration, namely The CDK library
(*http://cdk*.*sourceforge*.*net*/)  has R integration since few years -
it might help if you consult the code there. And AFAIK the author of
CDK-R code is on this list as well, so please do ask specific questions
on the development mailing list.
> 6) We are using random forests classification algorithm. By the
> Opentox model we should have access to this algorithm by an API. If
> that is the case then we can easily use that for development of our
> own model. However, as yet, it appears that no one has a algorithm/API
> for random forests in opentox. How should we proceed in this regard?
>
>   
The API was designed to be generic, and there is intentionally no
specific API for different type of machine learning algorithms.  The API
allows to pass parameters, that are specific to the algorithm  , plus a
training dataset.  If you could describe your workflow step by step, we
might try to map it into the current API and , if it is not sufficient,
extend the API.

Hope this helps,
Nina
> Looking fwd to your gudiance.
>
> Regards
>
> Sunil
>   <>
>
>
>
> Vedrin Jeliazkov wrote:
>
> Hi Sunil,
> 2009/12/21 sunil chawla <sunil at seascapelearning.com>:
>
>
> If there is some mechanism to give developers feedback/suggestions for
> improvements or course corrections after testing of submiited services - it
> would be helpful to us and bring quicker convergence.
>
>
> Well, there are already several such mechanisms in place and in active use:
> 1) Developer's own bug (or issue) tracker, which has to be linked from
> the webservice test instance description page:
> http://www.opentox.org/dev/testing/testtoxservices
> (most developers prefer this option so far)
> 2) OpenTox test results issue tracker:
> http://www.opentox.org/dev/testing/testresults/issues
> (this might be suitable for those who don't run their own issue tracker)
> 3) OT-Dev list, for any generic (e.g. API-related) issues.
> 4) Direct communication via e-mail or any other online collaboration
> tool for issues which are service/implementation- specific and need
> some more in depth discussion.
> Do you have something else in mind?
> Kind regards,
> Vedrin
> PS: Could you give us an update (preferably before tomorrow's meeting)
> on SL's progress with software development?
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>