You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@lucene.apache.org by Grant Ingersoll <gs...@apache.org> on 2009/04/07 22:08:05 UTC
[ANNOUNCE] Apache Mahout 0.1 Released
The Apache Lucene project is pleased to announce the release of Apache
Mahout 0.1.
Apache Mahout is a subproject of Apache Lucene with the goal of
delivering scalable
machine learning algorithm implementations under the Apache license.
The first public
release includes implementations for clustering, classification,
collaborative filtering and evolutionary programming.
Highlights include:
1. Taste Collaborative Filtering
2. Several distributed clustering implementations: k-Means, Fuzzy k-
Means, Dirchlet, Mean-Shift and Canopy
3. Distributed Naive Bayes and Complementary Naive Bayes
classification implementations
4. Distributed fitness function implementation for the Watchmaker
evolutionary programming library
5. Most implementations are built on top of Apache Hadoop (http://hadoop.apache.org
) for scalability
The release contents have been pushed out to the main Apache release
site and the m2 ibiblio sync repository.
Apache Mahout 0.1 is the project's first release and is focused on
establishing a baseline release while
attracting more contributors. Details can
be found in JIRA:
https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12310751&styleName=Html&version=12312976
Apache Mahout is available in source form from the following download
page:
http://www.apache.org/dyn/closer.cgi/lucene/mahout/0.1/mahout-0.1-project.tar.gz
Apache Mahout is also available for Maven 2 users via
the Central Maven Repositories:
http://repo1.maven.org/maven2/org/apache/mahout/
http://mirrors.ibiblio.org/pub/mirrors/maven2/org/apache/mahout/
When downloading from a mirror site, please remember to verify the
downloads
using signatures found on the Apache site:
http://www.apache.org/dist/lucene/mahout/KEYS
For more information on Apache Mahout, visit the project home page:
http://lucene.apache.org/mahout
Re: Interpolated Knn and SVD Recommender
Posted by Grant Ingersoll <gs...@apache.org>.
On Apr 11, 2009, at 2:46 AM, Sean Owen wrote:
>
> Let's do another round of patch-trading here to get it from 90% there
> to 99% there before committing. I don't disagree with the philosophy;
> perhaps it is just a matter of opinion on how long you need to iterate
> on a patch before it qualifies for SVN.
Totally agree, it's up to the committers job to make that decision on
when it is ready. I just wasn't sure if you were implying you wanted
to commit just to make it easier to develop on.
-Grant
Re: Interpolated Knn and SVD Recommender
Posted by Sean Owen <sr...@gmail.com>.
OK, yes I suppose I meant the patch looks OK to me to commit as-is,
maybe modulo stuff like copyright headers and tweaks here and there. I
didn't see how new code would break anyone's build (assuming it
compiles), and draft code can always be changed by others later; it's
a lot easier for me to review once it's in my source tree. SVN is of
course there to facilitate collaboration so I don't find it efficient
to trade patches manually -- of course we have the problem there that
two people can't iterate on it at once.
Let's do another round of patch-trading here to get it from 90% there
to 99% there before committing. I don't disagree with the philosophy;
perhaps it is just a matter of opinion on how long you need to iterate
on a patch before it qualifies for SVN.
On Sat, Apr 11, 2009 at 12:21 AM, Grant Ingersoll <gs...@apache.org> wrote:
>
> On Apr 10, 2009, at 7:02 PM, Sean Owen wrote:
>>
>> So that we can start collaborating, shall I go ahead and commit what
>> you have so far? If it's most of what needs to be done, let's do it.
>> If there are significant changes left, then it will be more convenient
>> to let you keep working on it rather than have to send many patches to
>> JIRA later.
>
>
> Not sure of your intent here, and we haven't really discussed it for the
> project, but generally speaking most the projects I've been on at the ASF
> are Review then Commit, not Commit then Review. It is pretty easy to upload
> patches, so I'd generally favor doing that as opposed to committing things
> you aren't sure on. That way, you don't break someone who happens to be on
> trunk. Plus, it gives others time to review.
>
> Cheers,
> Grant
>
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Grant Ingersoll <gs...@apache.org>.
Yes, you missed the "publish" part of it:
See step 8 of http://cwiki.apache.org/MAHOUT/howtoupdatethewebsite.html
I'll take care of it.
On Apr 14, 2009, at 5:11 PM, Sean Owen wrote:
> Javadoc is here brah - I think I updated the Resources page but
> perhaps did
> not properly publish the link?
>
> ---------- Forwarded message ----------
> From: "Grant Ingersoll" <gs...@apache.org>
> Date: Apr 10, 2009 8:31 PM
> Subject: Re: [ANNOUNCE] Apache Mahout 0.1 Released
> To: <ma...@lucene.apache.org>
>
> On Apr 10, 2009, at 3:10 PM, Sean Owen wrote: > OK copied everything
> to
> .../mahout/javadoc/core --...
> Once it propagates, it will be at
> http://lucene.apache.org/mahout/javadoc/core/
>
> So, I think all that is needed at this point is a pointer to them
> from the
> Resource section of the site.
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Grant Ingersoll <gs...@apache.org>.
I'll check.
On Apr 14, 2009, at 5:11 PM, Sean Owen wrote:
> Javadoc is here brah - I think I updated the Resources page but
> perhaps did
> not properly publish the link?
>
> ---------- Forwarded message ----------
> From: "Grant Ingersoll" <gs...@apache.org>
> Date: Apr 10, 2009 8:31 PM
> Subject: Re: [ANNOUNCE] Apache Mahout 0.1 Released
> To: <ma...@lucene.apache.org>
>
> On Apr 10, 2009, at 3:10 PM, Sean Owen wrote: > OK copied everything
> to
> .../mahout/javadoc/core --...
> Once it propagates, it will be at
> http://lucene.apache.org/mahout/javadoc/core/
>
> So, I think all that is needed at this point is a pointer to them
> from the
> Resource section of the site.
Fwd: Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Sean Owen <sr...@gmail.com>.
Javadoc is here brah - I think I updated the Resources page but perhaps did
not properly publish the link?
---------- Forwarded message ----------
From: "Grant Ingersoll" <gs...@apache.org>
Date: Apr 10, 2009 8:31 PM
Subject: Re: [ANNOUNCE] Apache Mahout 0.1 Released
To: <ma...@lucene.apache.org>
On Apr 10, 2009, at 3:10 PM, Sean Owen wrote: > OK copied everything to
.../mahout/javadoc/core --...
Once it propagates, it will be at
http://lucene.apache.org/mahout/javadoc/core/
So, I think all that is needed at this point is a pointer to them from the
Resource section of the site.
Re: Interpolated Knn and SVD Recommender
Posted by André Panisson <pa...@di.unito.it>.
Sean, I think that with the advice of Grant in mind, If you want, you
could make your modifications, adapt to the project format
specifications and create another patch, so I could review and make
little corrections. Almost everything is ready in this patch - the hard
work of the expectation-maximization SVD and the matrix factorizations
are done. It wold be missing only little corrections that I can add later.
André
Grant Ingersoll wrote:
>
> On Apr 10, 2009, at 7:02 PM, Sean Owen wrote:
>>
>> So that we can start collaborating, shall I go ahead and commit what
>> you have so far? If it's most of what needs to be done, let's do it.
>> If there are significant changes left, then it will be more convenient
>> to let you keep working on it rather than have to send many patches to
>> JIRA later.
>
>
> Not sure of your intent here, and we haven't really discussed it for
> the project, but generally speaking most the projects I've been on at
> the ASF are Review then Commit, not Commit then Review. It is pretty
> easy to upload patches, so I'd generally favor doing that as opposed
> to committing things you aren't sure on. That way, you don't break
> someone who happens to be on trunk. Plus, it gives others time to
> review.
>
> Cheers,
> Grant
>
Re: Interpolated Knn and SVD Recommender
Posted by Grant Ingersoll <gs...@apache.org>.
On Apr 10, 2009, at 7:02 PM, Sean Owen wrote:
>
> So that we can start collaborating, shall I go ahead and commit what
> you have so far? If it's most of what needs to be done, let's do it.
> If there are significant changes left, then it will be more convenient
> to let you keep working on it rather than have to send many patches to
> JIRA later.
Not sure of your intent here, and we haven't really discussed it for
the project, but generally speaking most the projects I've been on at
the ASF are Review then Commit, not Commit then Review. It is pretty
easy to upload patches, so I'd generally favor doing that as opposed
to committing things you aren't sure on. That way, you don't break
someone who happens to be on trunk. Plus, it gives others time to
review.
Cheers,
Grant
Re: Interpolated Knn and SVD Recommender
Posted by Sean Owen <sr...@gmail.com>.
On Fri, Apr 10, 2009 at 10:58 PM, André Panisson <pa...@di.unito.it> wrote:
> I uploaded on Jira the version I'm working with of the interpolated Knn
> and SVD Recommenders. There are still some fixes to make, and some
> modifications as adding documentation and license.
> Some work is still needed in the Knn implementation to follow exactly
> the paper description, but I'm already having very good results with it.
> These implementations are basically a reimplementation of the generic
> item-based recommender, and I don't know exactly how to implement the
> recommend() method (Sean, could you help on this?). The
It's probably mostly copy-and-paste from the original method -- should
be quite similar. I'll take a look later at whether it should be like
an item-based recommender or whether the framework needs new
abstractions for this type of recommender.
So that we can start collaborating, shall I go ahead and commit what
you have so far? If it's most of what needs to be done, let's do it.
If there are significant changes left, then it will be more convenient
to let you keep working on it rather than have to send many patches to
JIRA later.
Interpolated Knn and SVD Recommender
Posted by André Panisson <pa...@di.unito.it>.
Hi all,
I uploaded on Jira the version I'm working with of the interpolated Knn
and SVD Recommenders. There are still some fixes to make, and some
modifications as adding documentation and license.
Some work is still needed in the Knn implementation to follow exactly
the paper description, but I'm already having very good results with it.
These implementations are basically a reimplementation of the generic
item-based recommender, and I don't know exactly how to implement the
recommend() method (Sean, could you help on this?). The
estimatePreference methods are working well, and I had very good results
with the reduced GroupLens data model (100k preferences): RMSE 0.9168
for the knn interpolated recommender using the inverse of mean error as
similarity, and RMSE 0.9215 for the svd recommender with 22 features and
30 training steps. To compare with the SlopeOne recommender, for
example, I had a RMSE of 0.9357.
André
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Grant Ingersoll <gs...@apache.org>.
On Apr 10, 2009, at 3:10 PM, Sean Owen wrote:
> OK copied everything to .../mahout/javadoc/core -- last basic
> question, where does it show up publicly then?
Once it propagates, it will be at http://lucene.apache.org/mahout/javadoc/core/
So, I think all that is needed at this point is a pointer to them from
the Resource section of the site.
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Sean Owen <sr...@gmail.com>.
OK copied everything to .../mahout/javadoc/core -- last basic
question, where does it show up publicly then?
On Fri, Apr 10, 2009 at 7:41 PM, Grant Ingersoll <gs...@apache.org> wrote:
>
> On Apr 10, 2009, at 2:12 PM, Sean Owen wrote:
>
>> PS I added javadoc generation to the pom.xml and generated the
>> javadoc. Now where does it go, what is this mirroring directory? for
>> the moment I stashed the javadoc for core/ here:
>
> people.apache.org:/www/lucene.apache.org/mahout
>
> Under there, I would just add an api or javadocs directory that contains the
> Javadocs and then patch the site:
> http://cwiki.apache.org/MAHOUT/howtoupdatethewebsite.html
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Grant Ingersoll <gs...@apache.org>.
On Apr 10, 2009, at 2:12 PM, Sean Owen wrote:
> PS I added javadoc generation to the pom.xml and generated the
> javadoc. Now where does it go, what is this mirroring directory? for
> the moment I stashed the javadoc for core/ here:
people.apache.org:/www/lucene.apache.org/mahout
Under there, I would just add an api or javadocs directory that
contains the Javadocs and then patch the site: http://cwiki.apache.org/MAHOUT/howtoupdatethewebsite.html
>
>
> http://taste.sourceforge.net/javadoc/index.html
>
> (Also I officially retired the old project page at sourceforge:
> http://taste.sourceforge.net/)
+1. I'm glad Taste is a part of Mahout!
>
>
> On Apr 10, 2009 2:50 PM, "Grant Ingersoll" <gs...@apache.org>
> wrote:
>
> Nothing automatic that I'm aware of. We should only publish released
> versions. On Lucene we just copy them into the mirroring directory on
> people.a.o and then link to them from the site.
>
> -Grant
>
> On Apr 10, 2009, at 9:46 AM, Sean Owen wrote: > Thanks for that -- how
> do we go about getting jav...
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Sean Owen <sr...@gmail.com>.
PS I added javadoc generation to the pom.xml and generated the
javadoc. Now where does it go, what is this mirroring directory? for
the moment I stashed the javadoc for core/ here:
http://taste.sourceforge.net/javadoc/index.html
(Also I officially retired the old project page at sourceforge:
http://taste.sourceforge.net/)
On Apr 10, 2009 2:50 PM, "Grant Ingersoll" <gs...@apache.org> wrote:
Nothing automatic that I'm aware of. We should only publish released
versions. On Lucene we just copy them into the mirroring directory on
people.a.o and then link to them from the site.
-Grant
On Apr 10, 2009, at 9:46 AM, Sean Owen wrote: > Thanks for that -- how
do we go about getting jav...
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Grant Ingersoll <gs...@apache.org>.
Nothing automatic that I'm aware of. We should only publish released
versions. On Lucene we just copy them into the mirroring directory on
people.a.o and then link to them from the site.
-Grant
On Apr 10, 2009, at 9:46 AM, Sean Owen wrote:
> Thanks for that -- how do we go about getting javadoc up? is there
> anything automatic or special?
>
> On Fri, Apr 10, 2009 at 1:26 PM, Grant Ingersoll
> <gs...@apache.org> wrote:
>> FYI, I did all of these.
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Sean Owen <sr...@gmail.com>.
Thanks for that -- how do we go about getting javadoc up? is there
anything automatic or special?
On Fri, Apr 10, 2009 at 1:26 PM, Grant Ingersoll <gs...@apache.org> wrote:
> FYI, I did all of these.
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Grant Ingersoll <gs...@apache.org>.
FYI, I did all of these.
On Apr 8, 2009, at 7:31 AM, Grant Ingersoll wrote:
> Yes, please do. While your at it, could you add a news item on the
> front page?
>
> Thanks!
>
> We also should put the release java docs up. And, I need to make a
> branch from the tag.
>
> -Grant
>
> On Apr 8, 2009, at 7:13 AM, Sean Owen wrote:
>
>> This is great. May I / shall I update our project Resources page to
>> link to these?
>>
>> On Tue, Apr 7, 2009 at 9:08 PM, Grant Ingersoll
>> <gs...@apache.org> wrote:
>>> The Apache Lucene project is pleased to announce the release of
>>> Apache
>>> Mahout 0.1.
>>> Apache Mahout is a subproject of Apache Lucene with the goal of
>>> delivering
>>> scalable
>>> machine learning algorithm implementations under the Apache
>>> license. The
>>> first public
>>> release includes implementations for clustering, classification,
>>> collaborative filtering and evolutionary programming.
>
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Grant Ingersoll <gs...@apache.org>.
Yes, please do. While your at it, could you add a news item on the
front page?
Thanks!
We also should put the release java docs up. And, I need to make a
branch from the tag.
-Grant
On Apr 8, 2009, at 7:13 AM, Sean Owen wrote:
> This is great. May I / shall I update our project Resources page to
> link to these?
>
> On Tue, Apr 7, 2009 at 9:08 PM, Grant Ingersoll
> <gs...@apache.org> wrote:
>> The Apache Lucene project is pleased to announce the release of
>> Apache
>> Mahout 0.1.
>> Apache Mahout is a subproject of Apache Lucene with the goal of
>> delivering
>> scalable
>> machine learning algorithm implementations under the Apache
>> license. The
>> first public
>> release includes implementations for clustering, classification,
>> collaborative filtering and evolutionary programming.
Re: [ANNOUNCE] Apache Mahout 0.1 Released
Posted by Sean Owen <sr...@gmail.com>.
This is great. May I / shall I update our project Resources page to
link to these?
On Tue, Apr 7, 2009 at 9:08 PM, Grant Ingersoll <gs...@apache.org> wrote:
> The Apache Lucene project is pleased to announce the release of Apache
> Mahout 0.1.
> Apache Mahout is a subproject of Apache Lucene with the goal of delivering
> scalable
> machine learning algorithm implementations under the Apache license. The
> first public
> release includes implementations for clustering, classification,
> collaborative filtering and evolutionary programming.