rat_race
Member level 2
hi everyone
i have a set of complex valued vectors which i want to perform principal component analysis on. so my data matrix will be all the vectors putten as columns of the matrix beside each other where the number of components of each vector is 10 and the total number of vectos is 1000, so that my data matrix's size will be 10 X 1000.
now as the data are all complex valued, trivially the correlation/covariance matrices will also be complex valued (except for the main diagonal) and consequently eigenanalysis of such correlation matrix will lead to eigen vactors with complex valued elements. now , the article that i am trying to re-implement says in it that the eigenvectors of such analysis will be real but has not described how !
Now can anybody tell me how on earth am i supposed to get MEANINGFUL real valued eigenvectors out of a complex valued data. please help me or at least introduce me a reference.
Best
i have a set of complex valued vectors which i want to perform principal component analysis on. so my data matrix will be all the vectors putten as columns of the matrix beside each other where the number of components of each vector is 10 and the total number of vectos is 1000, so that my data matrix's size will be 10 X 1000.
now as the data are all complex valued, trivially the correlation/covariance matrices will also be complex valued (except for the main diagonal) and consequently eigenanalysis of such correlation matrix will lead to eigen vactors with complex valued elements. now , the article that i am trying to re-implement says in it that the eigenvectors of such analysis will be real but has not described how !
Now can anybody tell me how on earth am i supposed to get MEANINGFUL real valued eigenvectors out of a complex valued data. please help me or at least introduce me a reference.
Best
