This function is not yet fully documented. This is a transcript of the text-formatted help.
bbpca Principal component analysis [SIG,MU,U,V]=bbpca(X,L), or [SIG,MU,U,V]=bbpca(X,L,'row'), computes the L leading principal components of the data-matrix X, assuming the objects are arranged as rows in X. X can be a matrix or black-box operator. SIG is a vector of the singular values of the centered data-matrix. The eigenvalues of the covariance matrix is given by LAMBDA=SIG.^2. MU is the empirical mean of the objects, which is subtracted each object before the leading components are computed. U contains the eigenvectors of the empirical covariance matrix. The factor-scores matrix is given by U*diag(sig), while the factor loading matrix is given by V. The data-matrix can be written in terms of all principal components: X = U*diag(SIG)*V' + repmat(MU,size(X,1),1) [SIG,MU,U,V]=bbpca(X,L,'col') computes the PCA when the columns are arranged in columns. Assuming all principal components are computed, the data-matrix is given by: X = U*diag(SIG)*V' + repmat(MU,1,size(X,2)) The factor-scores matrix is given by diag(sig)*V, and the factor-loading matrix is U. bbpca uses bbsvds for the computation. As a last argument you may pass a struct to control this computation. Example: [sig,mu,U,V]=bbpca(X,100); See also bbsvds, BBSVDSOPT.