automatic commit

[folded-ctf.git] / loss_machine.h

diff --git a/loss_machine.h b/loss_machine.h
new file mode 100644 (file)
index 0000000..a293e8b
--- /dev/null
+++ b/loss_machine.h
@@ -0,0 +1,55 @@
+
+///////////////////////////////////////////////////////////////////////////
+// This program is free software: you can redistribute it and/or modify  //
+// it under the terms of the version 3 of the GNU General Public License //
+// as published by the Free Software Foundation.                         //
+//                                                                       //
+// This program is distributed in the hope that it will be useful, but   //
+// WITHOUT ANY WARRANTY; without even the implied warranty of            //
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU      //
+// General Public License for more details.                              //
+//                                                                       //
+// You should have received a copy of the GNU General Public License     //
+// along with this program. If not, see <http://www.gnu.org/licenses/>.  //
+//                                                                       //
+// Written by Francois Fleuret, (C) IDIAP                                //
+// Contact <francois.fleuret@idiap.ch> for comments & bug reports        //
+///////////////////////////////////////////////////////////////////////////
+
+#ifndef LOSS_MACHINE_H
+#define LOSS_MACHINE_H
+
+#include "misc.h"
+#include "sample_set.h"
+
+class LossMachine {
+  int _loss_type;
+
+public:
+  LossMachine(int loss_type);
+
+  void get_loss_derivatives(SampleSet *samples,
+                            scalar_t *responses,
+                            scalar_t *derivatives);
+
+  scalar_t loss(SampleSet *samples, scalar_t *responses);
+
+  scalar_t optimal_weight(SampleSet *sample_set,
+                          scalar_t *weak_learner_responses,
+                          scalar_t *current_responses);
+
+  // This method returns in sample_nb_occurences[k] the number of time
+  // the example k was sampled, and in sample_responses[k] the
+  // consistent response so that the overall loss remains the same. If
+  // allow_duplicates is set to 1, all samples will have an identical
+  // response (i.e. weight), but some may have more than one
+  // occurence. On the contrary, if allow_duplicates is 0, samples
+  // will all have only one occurence (or zero) but the responses may
+  // vary to account for the multiple sampling.
+
+  void subsample(int nb, scalar_t *labels, scalar_t *responses,
+                 int nb_to_sample, int *sample_nb_occurences, scalar_t *sample_responses,
+                 int allow_duplicates);
+};
+
+#endif