On Tue, Dec 13, 2011 at 07:31:36PM +0530, Jaidev Deshpande wrote:
> >> x_cent=zscore(x) % normalizing the data

Zscore centers and scales, right?

> >> [v,d]=eig(cov(x_cent))
> >> v=fliplr(v) % the eigenvectors need to be flipped because MATLAB returns 
> >> them in the descending order
> >> x_red = x_cent' * v(:,1:2)

> With decomposition.PCA I did the following (it's almost like the PCA
> vs LDA example in the docs):

> >>> pca=PCA(n_components=2)
> >>> x_red=pca.fit(x).transform(x)

> What's so different about the PCA implementation here in sklearn that
> I'm getting much more streamlined points in the scatter-plot?

PCA doesn't scale the input data.

G

------------------------------------------------------------------------------
Cloud Computing - Latest Buzzword or a Glimpse of the Future?
This paper surveys cloud computing today: What are the benefits? 
Why are businesses embracing it? What are its payoffs and pitfalls?
http://www.accelacomm.com/jaw/sdnl/114/51425149/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to