Minmax and Faq convergence of PARAFAC: Difference between pages

From Eigenvector Research Documentation Wiki
(Difference between pages)
Jump to navigation Jump to search
imported>Donal
No edit summary
 
imported>Lyle
No edit summary
 
Line 1: Line 1:
===Purpose===
===Issue:===


Scale rows or columns of a matrix to range from 0 to 1.
Convergence of PARAFAC. How much variation between models is expected a particular PARAFAC is fit multiple times with the same settings?


===Synopsis===
===Possible Solutions:===


:[xcorr,mins,maxs] = minmax(x,''options'');
Correctly converged models can vary in the loadings (e.g. permutation of components) but the fit should be exactly the same (e.g. as expressed by the sum of the squared residuals).


===Description===
If repeatedly fitted models are not identical in fit, it is an indication that your problem is very complex. It can be because you use too many components, because the real components are too similar within the signal-to-noise or because your data do not follow a low-rank trilinear model.


Scales rows (or columns) of a matrix (x) so each row (or column) has a minimum of 0 and a maximum of 1.  The option "mode" specifies whether rows or columns are scaled to have unity range. Optional input (options) is described below.


===Inputs===
'''Still having problems? Please contact our helpdesk at [mailto:helpdesk@eigenvector.com helpdesk@eigenvector.com]'''


* '''x''' = ''M'' by ''N'' matrix of data to be scaled (class "double" or "dataset").
[[Category:FAQ]]
 
===Options===
 
''options'' =  a structure array with the following fields:
 
* '''mode''': [ 1 ] dimension of data on which to calculate the minima and maxima for scaling. 1 = over rows (each row will have range [0,1]); 2 = over columns (each column will have range [0,1]). Default is 1.
 
===Outputs===
 
* '''xcorr''' = the scaled data (xcorr will be the same class as x)
 
* '''mins''' = vector of minima for each row (or column)
 
* '''maxs''' = vector of maxima for each row (or column)
 
 
===See Also===
 
[[normaliz]], [[preprocess]], [[snv]]

Latest revision as of 12:00, 5 December 2018

Issue:

Convergence of PARAFAC. How much variation between models is expected a particular PARAFAC is fit multiple times with the same settings?

Possible Solutions:

Correctly converged models can vary in the loadings (e.g. permutation of components) but the fit should be exactly the same (e.g. as expressed by the sum of the squared residuals).

If repeatedly fitted models are not identical in fit, it is an indication that your problem is very complex. It can be because you use too many components, because the real components are too similar within the signal-to-noise or because your data do not follow a low-rank trilinear model.


Still having problems? Please contact our helpdesk at helpdesk@eigenvector.com

Retrieved from "https://www.wiki.eigenvector.com/index.php?title=Minmax&oldid=10052"