You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@giraph.apache.org by Sebastian Schelter <ss...@apache.org> on 2013/03/12 08:50:04 UTC
Experimental Integration into Mahout
Hi fellow Giraphers,
on Mahout's dev@ mailinglist, there is a discussion about a potential
(experimental) future integration of Giraph. Might be good to hear your
opinion about that.
Best,
Sebastian
Re: Experimental Integration into Mahout
Posted by Gustavo Enrique Salazar Torres <gs...@ime.usp.br>.
+1 (examples could be used for tutorial purposes).
On Tue, Mar 12, 2013 at 4:47 PM, Mirko Kämpf <mi...@cloudera.com>wrote:
> +1
>
>
> On Tue, Mar 12, 2013 at 8:35 PM, Sebastian Schelter <ss...@apache.org>
> wrote:
>
> > +1
> >
> > On 12.03.2013 20:33, Jake Mannix wrote:
> > > +1 to this distinction
> > >
> > >
> > > On Tue, Mar 12, 2013 at 12:22 PM, Avery Ching <ac...@apache.org>
> wrote:
> > >
> > >> My thoughts were that examples would be like toy examples (not
> suitable
> > >> for production) that might be used, say, for unittesting.
> Applications
> > >> would be full-fledged ready to run applications that should have
> > perhaps a
> > >> supporting input format for actual data.
> > >>
> > >>
> > >> On 3/12/13 12:19 PM, Claudio Martella wrote:
> > >>
> > >>> My thoughts are: what is the difference then between the applications
> > and
> > >>> the examples package?
> > >>>
> > >>>
> > >>> On Tue, Mar 12, 2013 at 8:18 PM, Avery Ching <ac...@apache.org>
> > wrote:
> > >>>
> > >>> HI Nicolas,
> > >>>>
> > >>>> I'd be interested in seeing this algorithm implemented in Giraph. My
> > >>>> suggestion would be to put it directly in Giraph. We should create
> a
> > new
> > >>>> maven module called giraph-applications and then maybe something
> like
> > >>>> org.apache.giraph.application.****affinity package. You can see
> > >>>> org.apache.giraph.examples as an example. We should also move the
> > other
> > >>>> actual applications out of examples so that they are easier to run.
> > >>>> Anyone
> > >>>> else have any thoughts?
> > >>>>
> > >>>> Avery
> > >>>>
> > >>>>
> > >>>> On 3/12/13 12:12 PM, Nicolas Lalevée wrote:
> > >>>>
> > >>>> I am not sure what you think by integration, but I am implementing
> > the
> > >>>>> affinity propagation algorithm proposed by Frey and Dueck [1]. And
> > >>>>> Giraph
> > >>>>> is a perfect fit for the implementation. And I was wondering, if I
> > have
> > >>>>> the
> > >>>>> opportunity to contribute that implementation, where should I post
> > it ?
> > >>>>> Giraph since it is based on it ? Or Mahout along with the other
> > >>>>> clustering
> > >>>>> algorithm ?
> > >>>>>
> > >>>>> I just subscribed to mahout-dev@ in case some discussion starts
> > there
> > >>>>> too.
> > >>>>>
> > >>>>> Nicolas
> > >>>>>
> > >>>>> [1] http://www.psi.toronto.edu/****affinitypropagation/**<
> > http://www.psi.toronto.edu/**affinitypropagation/**>
> > >>>>> FreyDueckScience07.pdf<http://**www.psi.toronto.edu/**
> > >>>>> affinitypropagation/**FreyDueckScience07.pdf<
> > http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf>
> > >>>>>>
> > >>>>>
> > >>>>> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a
> > écrit :
> > >>>>>
> > >>>>> Hi fellow Giraphers,
> > >>>>>
> > >>>>>> on Mahout's dev@ mailinglist, there is a discussion about a
> > potential
> > >>>>>> (experimental) future integration of Giraph. Might be good to hear
> > your
> > >>>>>> opinion about that.
> > >>>>>>
> > >>>>>> Best,
> > >>>>>> Sebastian
> > >>>>>>
> > >>>>>>
> > >>>
> > >>
> > >
> > >
> >
> >
>
>
> --
> --
> Mirko Kämpf
>
> *Trainer* @ Cloudera
>
> tel: +49 *176 20 63 51 99*
> skype: *kamir1604*
> mirko@cloudera.com
>
--
Gustavo Salazar Torres
---------------------------------
- "When describing your own work, be humble and don’t use superlatives of
praise, either explicitly or implicitly, even if you are enthusiastic"
- Mathematical Writing, Donald Knuth
Re: Experimental Integration into Mahout
Posted by Mirko Kämpf <mi...@cloudera.com>.
+1
On Tue, Mar 12, 2013 at 8:35 PM, Sebastian Schelter <ss...@apache.org> wrote:
> +1
>
> On 12.03.2013 20:33, Jake Mannix wrote:
> > +1 to this distinction
> >
> >
> > On Tue, Mar 12, 2013 at 12:22 PM, Avery Ching <ac...@apache.org> wrote:
> >
> >> My thoughts were that examples would be like toy examples (not suitable
> >> for production) that might be used, say, for unittesting. Applications
> >> would be full-fledged ready to run applications that should have
> perhaps a
> >> supporting input format for actual data.
> >>
> >>
> >> On 3/12/13 12:19 PM, Claudio Martella wrote:
> >>
> >>> My thoughts are: what is the difference then between the applications
> and
> >>> the examples package?
> >>>
> >>>
> >>> On Tue, Mar 12, 2013 at 8:18 PM, Avery Ching <ac...@apache.org>
> wrote:
> >>>
> >>> HI Nicolas,
> >>>>
> >>>> I'd be interested in seeing this algorithm implemented in Giraph. My
> >>>> suggestion would be to put it directly in Giraph. We should create a
> new
> >>>> maven module called giraph-applications and then maybe something like
> >>>> org.apache.giraph.application.****affinity package. You can see
> >>>> org.apache.giraph.examples as an example. We should also move the
> other
> >>>> actual applications out of examples so that they are easier to run.
> >>>> Anyone
> >>>> else have any thoughts?
> >>>>
> >>>> Avery
> >>>>
> >>>>
> >>>> On 3/12/13 12:12 PM, Nicolas Lalevée wrote:
> >>>>
> >>>> I am not sure what you think by integration, but I am implementing
> the
> >>>>> affinity propagation algorithm proposed by Frey and Dueck [1]. And
> >>>>> Giraph
> >>>>> is a perfect fit for the implementation. And I was wondering, if I
> have
> >>>>> the
> >>>>> opportunity to contribute that implementation, where should I post
> it ?
> >>>>> Giraph since it is based on it ? Or Mahout along with the other
> >>>>> clustering
> >>>>> algorithm ?
> >>>>>
> >>>>> I just subscribed to mahout-dev@ in case some discussion starts
> there
> >>>>> too.
> >>>>>
> >>>>> Nicolas
> >>>>>
> >>>>> [1] http://www.psi.toronto.edu/****affinitypropagation/**<
> http://www.psi.toronto.edu/**affinitypropagation/**>
> >>>>> FreyDueckScience07.pdf<http://**www.psi.toronto.edu/**
> >>>>> affinitypropagation/**FreyDueckScience07.pdf<
> http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf>
> >>>>>>
> >>>>>
> >>>>> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a
> écrit :
> >>>>>
> >>>>> Hi fellow Giraphers,
> >>>>>
> >>>>>> on Mahout's dev@ mailinglist, there is a discussion about a
> potential
> >>>>>> (experimental) future integration of Giraph. Might be good to hear
> your
> >>>>>> opinion about that.
> >>>>>>
> >>>>>> Best,
> >>>>>> Sebastian
> >>>>>>
> >>>>>>
> >>>
> >>
> >
> >
>
>
--
--
Mirko Kämpf
*Trainer* @ Cloudera
tel: +49 *176 20 63 51 99*
skype: *kamir1604*
mirko@cloudera.com
Re: Experimental Integration into Mahout
Posted by Sebastian Schelter <ss...@apache.org>.
+1
On 12.03.2013 20:33, Jake Mannix wrote:
> +1 to this distinction
>
>
> On Tue, Mar 12, 2013 at 12:22 PM, Avery Ching <ac...@apache.org> wrote:
>
>> My thoughts were that examples would be like toy examples (not suitable
>> for production) that might be used, say, for unittesting. Applications
>> would be full-fledged ready to run applications that should have perhaps a
>> supporting input format for actual data.
>>
>>
>> On 3/12/13 12:19 PM, Claudio Martella wrote:
>>
>>> My thoughts are: what is the difference then between the applications and
>>> the examples package?
>>>
>>>
>>> On Tue, Mar 12, 2013 at 8:18 PM, Avery Ching <ac...@apache.org> wrote:
>>>
>>> HI Nicolas,
>>>>
>>>> I'd be interested in seeing this algorithm implemented in Giraph. My
>>>> suggestion would be to put it directly in Giraph. We should create a new
>>>> maven module called giraph-applications and then maybe something like
>>>> org.apache.giraph.application.****affinity package. You can see
>>>> org.apache.giraph.examples as an example. We should also move the other
>>>> actual applications out of examples so that they are easier to run.
>>>> Anyone
>>>> else have any thoughts?
>>>>
>>>> Avery
>>>>
>>>>
>>>> On 3/12/13 12:12 PM, Nicolas Lalevée wrote:
>>>>
>>>> I am not sure what you think by integration, but I am implementing the
>>>>> affinity propagation algorithm proposed by Frey and Dueck [1]. And
>>>>> Giraph
>>>>> is a perfect fit for the implementation. And I was wondering, if I have
>>>>> the
>>>>> opportunity to contribute that implementation, where should I post it ?
>>>>> Giraph since it is based on it ? Or Mahout along with the other
>>>>> clustering
>>>>> algorithm ?
>>>>>
>>>>> I just subscribed to mahout-dev@ in case some discussion starts there
>>>>> too.
>>>>>
>>>>> Nicolas
>>>>>
>>>>> [1] http://www.psi.toronto.edu/****affinitypropagation/**<http://www.psi.toronto.edu/**affinitypropagation/**>
>>>>> FreyDueckScience07.pdf<http://**www.psi.toronto.edu/**
>>>>> affinitypropagation/**FreyDueckScience07.pdf<http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf>
>>>>>>
>>>>>
>>>>> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a écrit :
>>>>>
>>>>> Hi fellow Giraphers,
>>>>>
>>>>>> on Mahout's dev@ mailinglist, there is a discussion about a potential
>>>>>> (experimental) future integration of Giraph. Might be good to hear your
>>>>>> opinion about that.
>>>>>>
>>>>>> Best,
>>>>>> Sebastian
>>>>>>
>>>>>>
>>>
>>
>
>
Re: Experimental Integration into Mahout
Posted by Jake Mannix <ja...@gmail.com>.
+1 to this distinction
On Tue, Mar 12, 2013 at 12:22 PM, Avery Ching <ac...@apache.org> wrote:
> My thoughts were that examples would be like toy examples (not suitable
> for production) that might be used, say, for unittesting. Applications
> would be full-fledged ready to run applications that should have perhaps a
> supporting input format for actual data.
>
>
> On 3/12/13 12:19 PM, Claudio Martella wrote:
>
>> My thoughts are: what is the difference then between the applications and
>> the examples package?
>>
>>
>> On Tue, Mar 12, 2013 at 8:18 PM, Avery Ching <ac...@apache.org> wrote:
>>
>> HI Nicolas,
>>>
>>> I'd be interested in seeing this algorithm implemented in Giraph. My
>>> suggestion would be to put it directly in Giraph. We should create a new
>>> maven module called giraph-applications and then maybe something like
>>> org.apache.giraph.application.****affinity package. You can see
>>> org.apache.giraph.examples as an example. We should also move the other
>>> actual applications out of examples so that they are easier to run.
>>> Anyone
>>> else have any thoughts?
>>>
>>> Avery
>>>
>>>
>>> On 3/12/13 12:12 PM, Nicolas Lalevée wrote:
>>>
>>> I am not sure what you think by integration, but I am implementing the
>>>> affinity propagation algorithm proposed by Frey and Dueck [1]. And
>>>> Giraph
>>>> is a perfect fit for the implementation. And I was wondering, if I have
>>>> the
>>>> opportunity to contribute that implementation, where should I post it ?
>>>> Giraph since it is based on it ? Or Mahout along with the other
>>>> clustering
>>>> algorithm ?
>>>>
>>>> I just subscribed to mahout-dev@ in case some discussion starts there
>>>> too.
>>>>
>>>> Nicolas
>>>>
>>>> [1] http://www.psi.toronto.edu/****affinitypropagation/**<http://www.psi.toronto.edu/**affinitypropagation/**>
>>>> FreyDueckScience07.pdf<http://**www.psi.toronto.edu/**
>>>> affinitypropagation/**FreyDueckScience07.pdf<http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf>
>>>> >
>>>>
>>>> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a écrit :
>>>>
>>>> Hi fellow Giraphers,
>>>>
>>>>> on Mahout's dev@ mailinglist, there is a discussion about a potential
>>>>> (experimental) future integration of Giraph. Might be good to hear your
>>>>> opinion about that.
>>>>>
>>>>> Best,
>>>>> Sebastian
>>>>>
>>>>>
>>
>
--
-jake
Re: Experimental Integration into Mahout
Posted by Avery Ching <ac...@apache.org>.
My thoughts were that examples would be like toy examples (not suitable
for production) that might be used, say, for unittesting. Applications
would be full-fledged ready to run applications that should have perhaps
a supporting input format for actual data.
On 3/12/13 12:19 PM, Claudio Martella wrote:
> My thoughts are: what is the difference then between the applications and
> the examples package?
>
>
> On Tue, Mar 12, 2013 at 8:18 PM, Avery Ching <ac...@apache.org> wrote:
>
>> HI Nicolas,
>>
>> I'd be interested in seeing this algorithm implemented in Giraph. My
>> suggestion would be to put it directly in Giraph. We should create a new
>> maven module called giraph-applications and then maybe something like
>> org.apache.giraph.application.**affinity package. You can see
>> org.apache.giraph.examples as an example. We should also move the other
>> actual applications out of examples so that they are easier to run. Anyone
>> else have any thoughts?
>>
>> Avery
>>
>>
>> On 3/12/13 12:12 PM, Nicolas Lalevée wrote:
>>
>>> I am not sure what you think by integration, but I am implementing the
>>> affinity propagation algorithm proposed by Frey and Dueck [1]. And Giraph
>>> is a perfect fit for the implementation. And I was wondering, if I have the
>>> opportunity to contribute that implementation, where should I post it ?
>>> Giraph since it is based on it ? Or Mahout along with the other clustering
>>> algorithm ?
>>>
>>> I just subscribed to mahout-dev@ in case some discussion starts there
>>> too.
>>>
>>> Nicolas
>>>
>>> [1] http://www.psi.toronto.edu/**affinitypropagation/**
>>> FreyDueckScience07.pdf<http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf>
>>>
>>> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a écrit :
>>>
>>> Hi fellow Giraphers,
>>>> on Mahout's dev@ mailinglist, there is a discussion about a potential
>>>> (experimental) future integration of Giraph. Might be good to hear your
>>>> opinion about that.
>>>>
>>>> Best,
>>>> Sebastian
>>>>
>
Re: Experimental Integration into Mahout
Posted by Claudio Martella <cl...@gmail.com>.
My thoughts are: what is the difference then between the applications and
the examples package?
On Tue, Mar 12, 2013 at 8:18 PM, Avery Ching <ac...@apache.org> wrote:
> HI Nicolas,
>
> I'd be interested in seeing this algorithm implemented in Giraph. My
> suggestion would be to put it directly in Giraph. We should create a new
> maven module called giraph-applications and then maybe something like
> org.apache.giraph.application.**affinity package. You can see
> org.apache.giraph.examples as an example. We should also move the other
> actual applications out of examples so that they are easier to run. Anyone
> else have any thoughts?
>
> Avery
>
>
> On 3/12/13 12:12 PM, Nicolas Lalevée wrote:
>
>> I am not sure what you think by integration, but I am implementing the
>> affinity propagation algorithm proposed by Frey and Dueck [1]. And Giraph
>> is a perfect fit for the implementation. And I was wondering, if I have the
>> opportunity to contribute that implementation, where should I post it ?
>> Giraph since it is based on it ? Or Mahout along with the other clustering
>> algorithm ?
>>
>> I just subscribed to mahout-dev@ in case some discussion starts there
>> too.
>>
>> Nicolas
>>
>> [1] http://www.psi.toronto.edu/**affinitypropagation/**
>> FreyDueckScience07.pdf<http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf>
>>
>> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a écrit :
>>
>> Hi fellow Giraphers,
>>>
>>> on Mahout's dev@ mailinglist, there is a discussion about a potential
>>> (experimental) future integration of Giraph. Might be good to hear your
>>> opinion about that.
>>>
>>> Best,
>>> Sebastian
>>>
>>
>
--
Claudio Martella
claudio.martella@gmail.com
Re: Experimental Integration into Mahout
Posted by Avery Ching <ac...@apache.org>.
HI Nicolas,
I'd be interested in seeing this algorithm implemented in Giraph. My
suggestion would be to put it directly in Giraph. We should create a
new maven module called giraph-applications and then maybe something
like org.apache.giraph.application.affinity package. You can see
org.apache.giraph.examples as an example. We should also move the other
actual applications out of examples so that they are easier to run.
Anyone else have any thoughts?
Avery
On 3/12/13 12:12 PM, Nicolas Lalevée wrote:
> I am not sure what you think by integration, but I am implementing the affinity propagation algorithm proposed by Frey and Dueck [1]. And Giraph is a perfect fit for the implementation. And I was wondering, if I have the opportunity to contribute that implementation, where should I post it ? Giraph since it is based on it ? Or Mahout along with the other clustering algorithm ?
>
> I just subscribed to mahout-dev@ in case some discussion starts there too.
>
> Nicolas
>
> [1] http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf
>
> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a écrit :
>
>> Hi fellow Giraphers,
>>
>> on Mahout's dev@ mailinglist, there is a discussion about a potential
>> (experimental) future integration of Giraph. Might be good to hear your
>> opinion about that.
>>
>> Best,
>> Sebastian
Re: Experimental Integration into Mahout
Posted by Sebastian Schelter <ss...@apache.org>.
The discussion is centered around potentially using Giraph as a BSP
framework for implementing machine learning algorithms. But it's really
only about the future (in terms of the next months/years). And it's
totally open what the result will be.
You already said that Giraph is a perfect fit, so I'd suggest thats the
place to contribute your code.
/s
On 12.03.2013 20:12, Nicolas Lalevée wrote:
> I am not sure what you think by integration, but I am implementing the affinity propagation algorithm proposed by Frey and Dueck [1]. And Giraph is a perfect fit for the implementation. And I was wondering, if I have the opportunity to contribute that implementation, where should I post it ? Giraph since it is based on it ? Or Mahout along with the other clustering algorithm ?
>
> I just subscribed to mahout-dev@ in case some discussion starts there too.
>
> Nicolas
>
> [1] http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf
>
> Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a écrit :
>
>> Hi fellow Giraphers,
>>
>> on Mahout's dev@ mailinglist, there is a discussion about a potential
>> (experimental) future integration of Giraph. Might be good to hear your
>> opinion about that.
>>
>> Best,
>> Sebastian
>
Re: Experimental Integration into Mahout
Posted by Nicolas Lalevée <ni...@hibnet.org>.
I am not sure what you think by integration, but I am implementing the affinity propagation algorithm proposed by Frey and Dueck [1]. And Giraph is a perfect fit for the implementation. And I was wondering, if I have the opportunity to contribute that implementation, where should I post it ? Giraph since it is based on it ? Or Mahout along with the other clustering algorithm ?
I just subscribed to mahout-dev@ in case some discussion starts there too.
Nicolas
[1] http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf
Le 12 mars 2013 à 08:50, Sebastian Schelter <ss...@apache.org> a écrit :
> Hi fellow Giraphers,
>
> on Mahout's dev@ mailinglist, there is a discussion about a potential
> (experimental) future integration of Giraph. Might be good to hear your
> opinion about that.
>
> Best,
> Sebastian