mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dmitriy Lyubimov <dlie...@gmail.com>
Subject Re: Mahout examples
Date Tue, 03 Apr 2012 20:20:10 GMT
Ziad,

you can look at SSVD/PCA doc on mahout wiki. It has an overview
section, which is not technically tutorial per se, but gives you
necessary background to use PCA/dimensionality reduction and LSA
fold-in/ similarity measure techniques. (well, perhaps not
similarities, not there).

 PCA/Dimensionality reduction  also just one step and is fully
supported by this single step assuming you have your data prepped
row-wise. I am not sure that there's much of tutorial to be had there,
again, it's one step as explained there. SVD-based LSA (and especially
LSI) is a somewhat more involved topic, but then again you can use
df/itf pipeline ( seq2sparse, etc) as defined in the book and then run
SSVD on the results to get desired LSA space. One detail that is often
desired for LSA fold-ins and which may not be (i am not sure)
described in sufficient detail in the book is the dictionary structure
(which probably still will have to be re-indexed in some more suitable
real-time format anyway).

SVD LSA is quite not the same as pLSA though (and i am not even sure
if Mahout supports pLSA pipeline. Mahout LDA is considered to be the
most successful push in that direction though.)



On Tue, Apr 3, 2012 at 12:53 PM, ziad kamel <ziad.kamel25@gmail.com> wrote:
> Some examples with new recommendations like pLSA, matrix factorization and PCA
>
> On Tue, Apr 3, 2012 at 1:35 PM, Saikat Kanjilal <sxk1969@hotmail.com> wrote:
>>
>> Ziad,Can you be more specific, the book has many different examples in various chapters,
what are some specific things you are looking for?Regards
>>
>>> Date: Tue, 3 Apr 2012 13:11:15 -0500
>>> Subject: Mahout examples
>>> From: ziad.kamel25@gmail.com
>>> To: user@mahout.apache.org
>>>
>>> Hi , I have the book mahout in Action , is there any other examples
>>> about recommendations ? Thanks
>>

Mime
View raw message