Right, even if their code doesn't seem to go beyond the proof of concept.
See the implementation of the datasets server-side :
https://github.com/psi-project/server/tree/master/private/datasets
A the moment there seem to be more momentum in distributed machine learning
(i.e. integration with popular distributed processing frameworks such as
Spark) than in federated approaches like this one. I guess it's mostly
because the formers have built-in primitives for (scalable) data movement
and task distribution. REST is just not so fit for this stuff IMHO.
It's a proven pattern to learn models on distributed systems and then
expose a prediction API through HTTP though. At least when you don't have
to pass much X data through the call :)
2014-03-17 11:57 GMT+01:00 Joel Nothman <[email protected]>:
> At least one project has exposed scikit-learn fitting and trained models
> as a web service: http://psi.cecs.anu.edu.au/
>
>
> On 17 March 2014 21:33, Eustache DIEMERT <[email protected]> wrote:
>
>> Hi Daniel,
>>
>> I don't know to which extent python's regular virtualenv [1] does not
>> provide the same kind of functionality as Docker ?
>>
>> For the REST API, it seems a good idea for executing models, but IMHO it
>> would be largely suboptimal to fit estimators on data passed over HTTP.
>>
>> Another info is that companies such as Continuum [2] already provide
>> different tools that seem to cover the most of your program (e.g. Conda,
>> Wakari).
>>
>> HTH,
>>
>> Eustache
>>
>> [1] http://virtualenv.readthedocs.org/en/latest/virtualenv.html
>>
>> [2] http://continuum.io/
>>
>>
>> 2014-03-13 10:26 GMT+01:00 Daniel Pyrathon <[email protected]>:
>>
>>> Hi,
>>>
>>> My name is Daniel Pyrathon. I am currently a third year BSc student in
>>> Computer Science at the University of Plymouth.
>>> Since my placement year I have been very interested in machine learning and
>>> AI in general. This has led me to self-learn some of the main models such
>>> as: Neural Networks (MLP, RBM), Naive Bayes classifiers, Genetic
>>> Algorithms, and I am currently learning RBMs because of my interest in deep
>>> learning. I love programming and I have always been active in the Open
>>> Source community (especially Python). In the past years I have attended the
>>> last 3 EuroPython conferences and I have been a staff member of the
>>> conference for the last 2 years.
>>>
>>> I am currently using Scikit (together with other libraries) for my
>>> dissertation project: a real-time Twitter sentiment analysis classifier. As
>>> my application is written in HTML5 and JS, I connect to Scikit using a
>>> RESTful API. The API I had built was so generic, it was then reused with
>>> other projects.
>>>
>>> The idea that has come out of this experience is to build a Machine
>>> Learning SaaS.
>>>
>>> The SaaS would be composed of:
>>>
>>>
>>> - A Docker image, deployable in one command. This will allow any
>>> developer who has an lxc compatible system to install and use Scikit in an
>>> isolated environment. As Docker containers are entirely isolated, this
>>> gives freedom to the programmer to update and deploy without having to
>>> think about the other components of his/her application.
>>> - A RESTful interface on top of Scikit, providing one universal way for
>>> programmers to train, test, and evaluate multiple machine learning
>>> techniques. This makes it very simple for programmers to integrate machine
>>> learning algorithms with existing components of their stack and scale
>>> massively.
>>>
>>>
>>>
>>> I have already pitched this idea to the Shogun community, and I am getting
>>> a lot of very interesting feedback. I would also like the Scikit community
>>> to let me know if they would be interested in this too, and give me
>>> possible suggestions.
>>>
>>>
>>>
>>> Thanks,
>>>
>>> Daniel Pyrathon
>>>
>>>
>>> --
>>> *************
>>>
>>> PirosB3
>>>
>>> https://github.com/PirosB3 <http://pirosb3.com>
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Learn Graph Databases - Download FREE O'Reilly Book
>>> "Graph Databases" is the definitive new guide to graph databases and
>>> their
>>> applications. Written by three acclaimed leaders in the field,
>>> this first edition is now available. Download your free book today!
>>> http://p.sf.net/sfu/13534_NeoTech
>>> _______________________________________________
>>> Scikit-learn-general mailing list
>>> [email protected]
>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>
>>>
>>
>>
>> ------------------------------------------------------------------------------
>> Learn Graph Databases - Download FREE O'Reilly Book
>> "Graph Databases" is the definitive new guide to graph databases and their
>> applications. Written by three acclaimed leaders in the field,
>> this first edition is now available. Download your free book today!
>> http://p.sf.net/sfu/13534_NeoTech
>> _______________________________________________
>> Scikit-learn-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>
>>
>
>
> ------------------------------------------------------------------------------
> Learn Graph Databases - Download FREE O'Reilly Book
> "Graph Databases" is the definitive new guide to graph databases and their
> applications. Written by three acclaimed leaders in the field,
> this first edition is now available. Download your free book today!
> http://p.sf.net/sfu/13534_NeoTech
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
------------------------------------------------------------------------------
Learn Graph Databases - Download FREE O'Reilly Book
"Graph Databases" is the definitive new guide to graph databases and their
applications. Written by three acclaimed leaders in the field,
this first edition is now available. Download your free book today!
http://p.sf.net/sfu/13534_NeoTech
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
