The Spider Objects

 
 
 
 
   
  Multivariate Regression via Stiefel Constraints      
   
 


   author : gb  
   ostart=[];        possible start output features  for optimization
   qstart=[];        possible start input features  for optimization
   sstart=[];        possible start weighting
    w0 initial guess: ostart * sstart *qstart'
  
   Final model : 
   o=[];
   s=[];
   q=[];
   w=[];
   
   Hyper params :
   gamma = 0;        regularization parameter 
   k=5;              allowed components 
   conv_eps=1e-3;    optimization criteria
  
   Example:
    Mrs  vs KPLS
    t=toyreg({'l=500','o=3','n=4'});
    d=gen(t);
    [r,a]=train( group({ mrs('k=2'), kpls('nroflatent=2')}),get(d,1:400));
    test(a,get(d,401:500),'quadratic_loss')

Reference : Multivariate Regression via Stiefel Constraints

Author : Gökhan H. BakIr , Arthur Gretton , Bernhard Schölkopf

Link : ----------------------------------------------------------------