Results for 1982-1-9 (cont.)


      TRIAL 1: Without wavelet transform i.e. direct
          For mapping   Model  --->  Observed

        learning MSE   testing MSE  training itns.
           0.480301     0.560473      10000
           0.429339     0.537917      20000
           0.413963     0.536395      30000         (*)

           TRIAL 2: Using wavelet transform
          For mapping   Model  --->  Observed
          (mappings in the transform spaces)

        learning MSE  testing MSE  training itns.

w1         0.033384     0.033139     5000
w2         0.011712     0.022229     5000
res        0.294392     0.317806    20000           
w2+res     0.455110     0.546747
w1+w2+res  0.533730     0.635152                    (*)

 TRIAL 3: For mapping   Model  --->  Observed res, and w2+res

        learning MSE   testing MSE  training itns.
res        0.306059     0.323072      40000         (*) 
w2+res     0.372577     0.461600      40000         (*)

(*) Results will be shown.