search for: candecomp

I wrote some functions for multiway CANDECOMP, i.e. for least squares fitting of a_{i_1\cdots i_m}\approx\sum_{s=1}^p x^1_{i_1s}x^1_{i_1s}\cdots x^m_{i_ms} with arrays of arbitrary dimension. Reminded me of the good old APL days. I could not find this in the archives, but if it's already there, I would appreciate if someone let me know...

...to decompose a tensor (array) of any order, as a generalisation of SVD also supporting non-identity metrics and penalisations. 2-way SVD with these extensions is also available. The package includes also some other multiway methods: PCAn (Tucker-n) and PARAFAC/CANDECOMP with these extensions. please send comments + looking for nice not too big multi-arrays for the next release demos Didier -- Didier G. Leibovici didier at fmrib.ox.ac.uk +44 (0)1865 222 739 Image Analysis Group fax:+44 (0)1865 222 717 Oxford University, Centre For...

...to decompose a tensor (array) of any order, as a generalisation of SVD also supporting non-identity metrics and penalisations. 2-way SVD with these extensions is also available. The package includes also some other multiway methods: PCAn (Tucker-n) and PARAFAC/CANDECOMP with these extensions. please send comments + looking for nice not too big multi-arrays for the next release demos Didier -- Didier G. Leibovici didier at fmrib.ox.ac.uk +44 (0)1865 222 739 Image Analysis Group fax:+44 (0)1865 222 717 Oxford University, Centre For...

...PTAK and in particular the command Candpara to perform the Parafac factorizationor of a tensor. The results are not encouraging as I expected, I'm starting a phase of analysis to see if there are errors. I pose a question and I hope you can help me. The command to run the factorization is: ## CANDECOMP/PARAFAC results<- CANDPARA(data_matrix, dim=3) summary(results) U<-results[[1]]$v V<-results[[2]]$v W<-results[[3]]$v data_matrix is a tensor of 943x1682x4. what I want understand is: U, V, W, are really the three factors that I should get from factorization? I hope someone can help...