2012/1/23 Dimitrios Pritsos <[email protected]>:
> On 01/23/2012 02:20 PM, Lars Buitinck wrote:
>> 2012/1/23 Dimitrios Pritsos<[email protected]>:
>>> On 01/23/2012 12:24 PM, Olivier Grisel wrote:
>>>> BTW: what is the structure of you data in PyTables? Is is mapped to a
>>>> scipy.sparse Compressed Sparse Row datastructure? How many features do
>>>> you have in your dataset?
>>> The training data are in a EArray (Compressed per row due to lots of
>>> zeros).
>>> I have 34000 Samples and the length of my Dictionary depending on the
>>> Training Set is about 1,500,000.
>>> However, using about 30,000 features seems satisfactory for a
>>> proof-of-concept case. However the samples needs to be approximately
>>> about 30-50k.
>> That would be doable. 30k features × 50k samples in a CSR matrix with
>> dtype=float32, assuming it's 90% zeros (a pessimistic guess for topic
>> spotting) would take just over 2GB.
>>
> I will give it a try however in some of my tests had a memory management
> problem. As I can recall it was mostly because of numpy function that
> might ask from pyTable to load every thing in main men. I guess some
> loops and some slicing might solve the problem.
>
> However I fist try to figure out how to use linear_model.SGDClassifier
> which it suppose to be capable to be trained in stages. Plus since I am
> using Linear Kernel it won't effect my results.
>
> Still I will give a try to the Sparse structure.

BTW, if you find a way to load your data into a
scipy.sparse.csr_matrix that fits in memory at once then you don't
need to bother with the `partial_fit` method of SGDClassifier. Just
use the regular fit method and you will be fine.

-- 
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel

------------------------------------------------------------------------------
Try before you buy = See our experts in action!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-dev2
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to