mloc_busca_valida

 MLOC_BUSCA_VALIDA grid search for the best of multiple models with validation

        [mloc_best,LogData] = mloc_busca_valida(mloc,yt,ut,pt,yv,uv,pv,ny,nu,bias,sigma)

   mloc: object containing a SOM with local models
     yt: outputs (training)
     ut: inputs (training)
     pt: selectors of dynamics (training)
     yv: outputs (validation)
     uv: inputs (validation)
     pt: selectors of dyanamics (validation)
     ny: vector with candidate AR orders (outputs delays) for grid search
     nu: vector with candidate X  orders (inputs  delays) for grid search
   bias: 0 no bias, 1 with bias
  sigma: vector with candidate local model widths for grid search

  DESCRIPTION:
     Train mloc ARX models many times using all combination of 
     candidate parameters ny, nu, sigma.

     Each mloc is simulated for validation data and 
     the simulated outputs are compared against the true validation outputs 
     Finally, it returns the mloc structure with best (smallest) AIC

   SEE ALSO
      MLOC_SOM, MLOC_LOC, MLOC_SIM, MLOC_PLANOS, MLOC_COMPARA