From: r i t a on 25 Mar 2007 12:18 Hi! I'm trying to use PCA in a pattern recognition problem but I don't know how to finish it. I'm using the function pca from STPRTOOL, which has the form, model=pca(input) after do this I have the model with the eigenvalues, but I don't have the eigenvectors. I don't know how I'm going to construct the final data with the choosen eigenvalues. Does anybody can help me? I have no idea how can I solve this problem. Rita
From: Yang Zhang on 25 Mar 2007 18:36 r i t a wrote: > > > Hi! > I'm trying to use PCA in a pattern recognition problem but I don't > know how to finish it. > > I'm using the function pca from STPRTOOL, which has the form, > > model=pca(input) > > after do this I have the model with the eigenvalues, but I don't > have > the eigenvectors. > I don't know how I'm going to construct the final data with the > choosen eigenvalues. > Does anybody can help me? > I have no idea how can I solve this problem. > > Rita > > Rita, I did not use any available functions but the idea for pattern recognition is: x - data matrix with zero mean and unit variance [u e v] = svd(x), v is the eigenvector matrix and you should select the leading ones according to eigenvalue (e matrix). After you get the model (eigenvalue and vectors), use it to project your new data. Hope this helps, Yang
From: SeBy on 26 Mar 2007 01:59 Have a look to an excellent reading of Kevin Murphy : <http://www.cs.ubc.ca/~murphyk/Teaching/Stat406-Spring07/reading/pcaHandout.pdf> ou will find material at <http://www.cs.ubc.ca/~murphyk/Teaching/Stat406-Spring07/index.html> Regards, SeBy Yang Zhang wrote: > > > r i t a wrote: >> >> >> Hi! >> I'm trying to use PCA in a pattern recognition problem but I > don't >> know how to finish it. >> >> I'm using the function pca from STPRTOOL, which has the form, >> >> model=pca(input) >> >> after do this I have the model with the eigenvalues, but I don't >> have >> the eigenvectors. >> I don't know how I'm going to construct the final data with the >> choosen eigenvalues. >> Does anybody can help me? >> I have no idea how can I solve this problem. >> >> Rita >> >> > > Rita, > > I did not use any available functions but the idea for pattern > recognition is: > > x - data matrix with zero mean and unit variance > > [u e v] = svd(x), v is the eigenvector matrix and you should select > the leading ones according to eigenvalue (e matrix). > > After you get the model (eigenvalue and vectors), use it to project > your new data. > > Hope this helps, > > Yang
|
Pages: 1 Prev: reading focal length from EXIF data Next: SIMULINK: Error using ==> RTW.genMakefileAndBuild |