Hi Dmitriy,
Just a few comments:
--the computed factors are approximate A \approx U\SigmaV^{T}
-- the projection steps seemed transposed to me but they are consistent
throughout ie.
(2) \tilde{u} = \tilde{c}_{r} V \Sigma^{-1}
p. 3: transpose \xi to emphasize row vector
- 'mean of all rows' is a bit misleading, \xi entries are the mean of each
column (column-wise mean as you state below)
- dimention -> dimension
I haven't code dived into the new pca code to be familiar with it so the
above comments are just picky notational stuff. I did however, do some
extensive analysis on the standard decomposition part (as of 0.6 SNAPSHOT)
which can be found here
http://amath.colorado.edu/faculty/martinss/Pubs/2012_halko_dissertation.pdf
(starting page 139)
Its a beast of a read so I can concisely discuss a few points perhaps on a
different thread.
On Wed, Feb 22, 2012 at 6:08 PM, Dmitriy Lyubimov <[email protected]> wrote:
> Could somebody please review SSVD command line usage doc before i
> update it on wiki for inclusion of complete nonsense, in particular,
> section $3 where it does overview of PCA and dimensionality reduction
> techniques? Here's the SSVD CLI doc:
>
>
> https://github.com/dlyubimov/mahout-commits/blob/ssvd-docs/SSVD-CLI.pdf?raw=true
>
>
> Thanks.
> -D
>
> On Wed, Feb 22, 2012 at 4:53 PM, Dmitriy Lyubimov <[email protected]>
> wrote:
> > Hi,
> > working on PCA section in SSVD usage .
> >
> > Just to confirm, if we run and svd over input with mean subtracted,
> > then U matrix presents original data points converted to PCA space,
> > right?
> >
> > thanks.
> > -d
>
