create a single training function for easy external access

7 files changed, 92 insertions, 40 deletions

diff --git a/Makefile b/Makefile
index d281bb4..4b4fbb2 100644
--- a/Makefile
+++ b/Makefile
@@ -47,7 +47,8 @@ lib/libgensvm.a: \
 	src/gensvm_strutil.o \
 	src/gensvm_sv.o \
 	src/gensvm_task.o \
-	src/gensvm_timer.o
+	src/gensvm_timer.o \
+	src/gensvm_train.o
 	@ar rcs lib/libgensvm.a \
 		src/gensvm_base.o \
 		src/gensvm_cmdarg.o \
@@ -67,7 +68,8 @@ lib/libgensvm.a: \
 		src/gensvm_strutil.o \
 		src/gensvm_sv.o \
 		src/gensvm_task.o \
-		src/gensvm_timer.o
+		src/gensvm_timer.o \
+		src/gensvm_train.o
 	@echo libgensvm.a...
 
 gensvm: src/GenSVMtraintest.c lib/libgensvm.a
diff --git a/include/gensvm_optimize.h b/include/gensvm_optimize.h
index 7a23bdb..1041b72 100644
--- a/include/gensvm_optimize.h
+++ b/include/gensvm_optimize.h
@@ -10,8 +10,8 @@
  *
  */
 
-#ifndef GENSVM_TRAIN_H
-#define GENSVM_TRAIN_H
+#ifndef GENSVM_OPTIMIZE_H
+#define GENSVM_OPTIMIZE_H
 
 #include "gensvm_sv.h"
 #include "gensvm_print.h"
diff --git a/include/gensvm_train.h b/include/gensvm_train.h
new file mode 100644
index 0000000..c410cef
--- /dev/null
+++ b/include/gensvm_train.h
@@ -0,0 +1,21 @@
+/**
+ * @file gensvm_train.h
+ * @author Gertjan van den Burg
+ * @date May, 2016
+ * @brief Header file for gensvm_train.c
+ *
+ */
+
+#ifndef GENSVM_TRAIN_H
+#define GENSVM_TRAIN_H
+
+// includes
+#include "gensvm_init.h"
+#include "gensvm_kernel.h"
+#include "gensvm_optimize.h"
+
+// function declarations
+void gensvm_train(struct GenModel *model, struct GenData *data,
+		struct GenModel *seed_model);
+
+#endif
diff --git a/src/GenSVMtraintest.c b/src/GenSVMtraintest.c
index 47cc900..73c0ce9 100644
--- a/src/GenSVMtraintest.c
+++ b/src/GenSVMtraintest.c
@@ -12,8 +12,7 @@
 
 #include "gensvm_cmdarg.h"
 #include "gensvm_io.h"
-#include "gensvm_init.h"
-#include "gensvm_optimize.h"
+#include "gensvm_train.h"
 #include "gensvm_pred.h"
 
 #define MINARGS 2
@@ -54,10 +53,8 @@ void exit_with_help()
 	exit(0);
 }
 
-
 int main(int argc, char **argv)
 {
-	bool with_test = false;
 	long i, *predy = NULL;
 	double performance;
 
@@ -68,6 +65,7 @@ int main(int argc, char **argv)
 	     *prediction_outputfile = NULL;
 
 	struct GenModel *model = gensvm_init_model();
+	struct GenModel *seed_model = NULL;
 	struct GenData *traindata = gensvm_init_data();
 	struct GenData *testdata = gensvm_init_data();
 
@@ -77,53 +75,29 @@ int main(int argc, char **argv)
 	parse_command_line(argc, argv, model, &model_inputfile,
 		       	&training_inputfile, &testing_inputfile,
 		       	&model_outputfile, &prediction_outputfile);
-	if (testing_inputfile != NULL)
-		with_test = true;
 
 	// read data from files
 	gensvm_read_data(traindata, training_inputfile);
-	if (with_test)
-		gensvm_read_data(testdata, testing_inputfile);
-
-	// copy dataset parameters to model
-	model->n = traindata->n;
-	model->m = traindata->m;
-	model->K = traindata->K;
 	model->data_file = training_inputfile;
 
-	// allocate model
-	gensvm_allocate_model(model);
-
-	// run pre/post processing in case of kernels
-	gensvm_kernel_preprocess(model, traindata);
-	if (with_test)
-		gensvm_kernel_postprocess(model, traindata, testdata);
-
-	// reallocate model in case of kernel dimension reduction
-	gensvm_reallocate_model(model, traindata->n, traindata->r);
-
-	// initialize weights
-	gensvm_initialize_weights(traindata, model);
-
 	// seed the random number generator
 	srand(time(NULL));
 
 	// load a seed model from file if it is specified
 	if (gensvm_check_argv_eq(argc, argv, "-s")) {
-		struct GenModel *seed_model = gensvm_init_model();
+		seed_model = gensvm_init_model();
 		gensvm_read_model(seed_model, model_inputfile);
-		gensvm_init_V(seed_model, model, traindata);
-		gensvm_free_model(seed_model);
-	} else {
-		gensvm_init_V(NULL, model, traindata);
 	}
 
-	// start training
-	gensvm_optimize(model, traindata);
+	// train the GenSVM model
+	gensvm_train(model, traindata, seed_model);
 
 	// if we also have a test set, predict labels and write to predictions
 	// to an output file if specified
-	if (with_test) {
+	if (testing_inputfile != NULL) {
+		gensvm_read_data(testdata, testing_inputfile);
+		gensvm_kernel_postprocess(model, traindata, testdata);
+
 		// predict labels
 		predy = Calloc(long, testdata->n);
 		gensvm_predict_labels(testdata, model, predy);
@@ -154,6 +128,7 @@ int main(int argc, char **argv)
 
 	// free everything
 	gensvm_free_model(model);
+	gensvm_free_model(seed_model);
 	gensvm_free_data(traindata);
 	gensvm_free_data(testdata);
 	free(training_inputfile);
diff --git a/src/gensvm_base.c b/src/gensvm_base.c
index eddef5c..568f19a 100644
--- a/src/gensvm_base.c
+++ b/src/gensvm_base.c
@@ -194,6 +194,9 @@ void gensvm_reallocate_model(struct GenModel *model, long n, long m)
  */
 void gensvm_free_model(struct GenModel *model)
 {
+	if (model == NULL)
+		return;
+
 	free(model->W);
 	free(model->t);
 	free(model->V);
diff --git a/src/gensvm_optimize.c b/src/gensvm_optimize.c
index 70b3620..464815d 100644
--- a/src/gensvm_optimize.c
+++ b/src/gensvm_optimize.c
@@ -1,5 +1,5 @@
 /**
- * @file gensvm_train.c
+ * @file gensvm_optimize.c
  * @author Gertjan van den Burg
  * @date August 9, 2013
  * @brief Main functions for training the GenSVM solution.
diff --git a/src/gensvm_train.c b/src/gensvm_train.c
new file mode 100644
index 0000000..4c0f332
--- /dev/null
+++ b/src/gensvm_train.c
@@ -0,0 +1,51 @@
+/**
+ * @file gensvm_train.c
+ * @author Gertjan van den Burg
+ * @date May, 2016
+ * @brief Main function for training a GenSVM model.
+ *
+ */
+
+#include "gensvm_train.h"
+
+/**
+ * @brief Utility function for training a GenSVM model
+ *
+ * @details
+ * This function organizes model allocation, kernel preprocessing, instance 
+ * weight initialization, and model training. It is the function that should 
+ * be used for training a single GenSVM model. Note that optionally a seed 
+ * model can be passed to the function to seed the V matrix with. When no such 
+ * model is used this parameter should be set to NULL.
+ *
+ * @param[in] 	model 		a GenModel instance
+ * @param[in] 	data 		a GenData instance with the training data
+ * @param[in] 	seed_model 	an optional GenModel to seed the V matrix
+ *
+ */
+void gensvm_train(struct GenModel *model, struct GenData *data,
+		struct GenModel *seed_model)
+{
+	// copy dataset parameters to model
+	model->n = data->n;
+	model->m = data->m;
+	model->K = data->K;
+
+	// initialize the V matrix (potentially with a seed model)
+	gensvm_init_V(seed_model, model, data);
+
+	// allocate model
+	gensvm_allocate_model(model);
+
+	// preprocess kernel
+	gensvm_kernel_preprocess(model, data);
+
+	// reallocate model for kernels
+	gensvm_reallocate_model(model, data->n, data->r);
+
+	// initialize weights
+	gensvm_initialize_weights(data, model);
+
+	// start training
+	gensvm_optimize(model, data);
+}