As far as I understand, Deepak was trying to obtain the same results with
both methods. His svd_t method centered along the feature axis, but he also
fed it with X.T, making it center along samples again. The surprising part
was that the last singular vectors did not correspond, and if I am not
mistaken, this is due to the centering operation and multiplicity of the
0-eigenspace. So what you are trying to do is slightly different: Using
svd_t on X or using pca on X.T.
Michael
On Sat, Aug 2, 2014 at 9:32 PM, Adam Hughes <[email protected]> wrote:
> Just to give an example, here's how I want to apply PCA to spectroscopy
> experiments:
>
> http://pdf.lookchem.com/pdf/32/f21930e7-09dc-4f70-ba24-f1edf8fdf02e.pdf
>
>
> On Sat, Aug 2, 2014 at 3:28 PM, Adam Hughes <[email protected]>
> wrote:
>
>> Deepak,
>>
>> Thanks for getting this discussion started. I also had some similar
>> concerns. There are some new applications in spectroscopy that use the
>> principle components on spectral data; however, in those systems, the
>> centering is done along the feature axis. Thus, my dataset is shaped like
>> your svd_T() example. This really helps me keep things straight, so I'll
>> probably put a reference to this thread in any PCA tutorials I put into my
>> spectroscopy package. Thanks also to the rest of the participants in this
>> thread for explaining some of this nuance that is often taken for granted
>> in PCA.
>>
>> I read through the thread, and am still not 100% sure what the best
>> approach for the SVD_T() case is. For data of shape samples X features,
>> and we want to do PCA along the features, is SVD_T() sufficient, or should
>> I wait for the incremental PCA PR?
>>
>>
>> On Thu, Jul 31, 2014 at 11:27 AM, Deepak Pandian <
>> [email protected]> wrote:
>>
>>> On Thu, Jul 31, 2014 at 8:50 PM, Michael Eickenberg
>>> <[email protected]> wrote:
>>> > Coming soon :) https://github.com/scikit-learn/scikit-learn/pull/3285
>>>
>>> That looks cool. I will look at it.
>>>
>>>
>>>
>>> --
>>> With Regards,
>>> Deepak Pandian
>>> "Deconstructing world one piece at a time"
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Infragistics Professional
>>> Build stunning WinForms apps today!
>>> Reboot your WinForms applications with our WinForms controls.
>>> Build a bridge from your legacy apps to the future.
>>>
>>> http://pubads.g.doubleclick.net/gampad/clk?id=153845071&iu=/4140/ostg.clktrk
>>> _______________________________________________
>>> Scikit-learn-general mailing list
>>> [email protected]
>>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>>
>>
>>
>
>
> ------------------------------------------------------------------------------
> Want fast and easy access to all the code in your enterprise? Index and
> search up to 200,000 lines of code with a free copy of Black Duck
> Code Sight - the same software that powers the world's largest code
> search on Ohloh, the Black Duck Open Hub! Try it now.
> http://p.sf.net/sfu/bds
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
------------------------------------------------------------------------------
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
