# svd and eigh Classic List Threaded 4 messages Open this post in threaded view
|

## svd and eigh

 I am trying out the eigenvectors related  functions in numpy.linalg.I came across some portions where i have doubts. 1). i have an array X if i calculate L=dot(X,X.transpose()) can L be called the covariance matrix of X?I read so in a paper by Turk&Pentland(equation 3 i think) can someone clarify this ? 2). i tried to find eigenvectors using svd() and eigh() functions evects1,evals,vt=svd(L,0) evals,evects2=eigh(L) and sorted both evects1 and evects2 in the descending order of their evals here i find that evects1 and evects2 have same values but some of the values differ in their signs.why is this? can anyone explain tia W _______________________________________________ Numpy-discussion mailing list [hidden email] http://projects.scipy.org/mailman/listinfo/numpy-discussion
Open this post in threaded view
|

## Re: svd and eigh

 On Fri, 2 May 2008 23:34:19 -0700 (PDT)   wilson <[hidden email]> wrote: > I am trying out the eigenvectors related  functions in >numpy.linalg.I > came across some portions where i have doubts. > 1). > i have an array X > if i calculate L=dot(X,X.transpose()) > can L be called the covariance matrix of X?I read so in >a paper by > Turk&Pentland(equation 3 i think) > can someone clarify this ? > > 2). > i tried to find eigenvectors using svd() and eigh() >functions > evects1,evals,vt=svd(L,0) > > evals,evects2=eigh(L) > > and sorted both evects1 and evects2 in the descending >order of their > evals > here i find that evects1 and evects2 have same values >but some of the > values differ in their signs.why is this? > > can anyone explain > tia > W > _______________________________________________ > Numpy-discussion mailing list > [hidden email] > http://projects.scipy.org/mailman/listinfo/numpy-discussionhttp://en.wikipedia.org/wiki/Singular_value_decompositionhttp://en.wikipedia.org/wiki/Eigenvectorhttp://en.wikipedia.org/wiki/Covariance_matrixHTH,        Nils _______________________________________________ Numpy-discussion mailing list [hidden email] http://projects.scipy.org/mailman/listinfo/numpy-discussion