Changeset 6718

Timestamp:

09/08/11 11:29:26 (13 years ago)

Author:

gkronber

Message:

#1474 fixed bug in alglib that resulting in random models when using the early stopping training method for multi layer perceptrons.

File:

• trunk/sources/HeuristicLab.ExtLibs/HeuristicLab.ALGLIB/3.1.0/ALGLIB-3.1.0/dataanalysis.cs (modified) (1 diff)

trunk/sources/HeuristicLab.ExtLibs/HeuristicLab.ALGLIB/3.1.0/ALGLIB-3.1.0/dataanalysis.cs

@@ -15269 +15269 @@
                     if( state.xupdated )
                     {
-                        for(i_=0; i_<=wcount-1;i_++)
-                        {
-                            network.weights[i_] = w[i_];
+                        minlbfgs.minlbfgsresults(state, ref w, internalrep);
+                        for (i_ = 0; i_ <= wcount - 1; i_++) {
+                          network.weights[i_] = w[i_];
                         }
                         e = mlpbase.mlperror(network, valxy, valsize);