work on reallocation of model and work on bugfix free

7 files changed, 105 insertions, 47 deletions

diff --git a/include/msvmmaj_train_dataset.h b/include/msvmmaj_train_dataset.h
index 2afcde8..0889626 100644
--- a/include/msvmmaj_train_dataset.h
+++ b/include/msvmmaj_train_dataset.h
@@ -135,7 +135,7 @@ double cross_validation(struct MajModel *model, struct MajData *data,
 void make_model_from_task(struct Task *task, struct MajModel *model);
 void copy_model(struct MajModel *from, struct MajModel *to);
 
-void msvmmaj_reallocate_model(struct MajModel *model, long n);
+void msvmmaj_reallocate_model(struct MajModel *model, long n, long m);
 
 void print_progress_string(struct Task *task, long N);
 #endif
diff --git a/src/libMSVMMaj.c b/src/libMSVMMaj.c
index a0bef97..65ff33d 100644
--- a/src/libMSVMMaj.c
+++ b/src/libMSVMMaj.c
@@ -126,7 +126,8 @@ void msvmmaj_simplex_diff(struct MajModel *model, struct MajData *data)
  * 				calculated with MajModel::V.
  *
  */
-void msvmmaj_calculate_errors(struct MajModel *model, struct MajData *data, double *ZV)
+void msvmmaj_calculate_errors(struct MajModel *model, struct MajData *data,
+	       	double *ZV)
 {
 	long i, j, k;
 	double a, value;
@@ -156,7 +157,8 @@ void msvmmaj_calculate_errors(struct MajModel *model, struct MajData *data, doub
 		for (j=0; j<K-1; j++) {
 			a = matrix_get(ZV, K-1, i, j);
 			for (k=0; k<K; k++) {
-				value = a * matrix3_get(model->UU, K-1, K, i, j, k);
+				value = a * matrix3_get(model->UU, K-1, K, i,
+					       	j, k);
 				matrix_add(model->Q, K, i, k, value);
 			}
 		}
diff --git a/src/msvmmaj_kernel.c b/src/msvmmaj_kernel.c
index a2c1193..edcd68f 100644
--- a/src/msvmmaj_kernel.c
+++ b/src/msvmmaj_kernel.c
@@ -118,7 +118,6 @@ void msvmmaj_make_kernel(struct MajModel *model, struct MajData *data)
 	long num_eigen = msvmmaj_make_eigen(K, n, &P, &Sigma);
 	//printf("num eigen: %li\n", num_eigen);
 	data->m = num_eigen;
-	model->m = num_eigen;
 
 	// copy eigendecomp to data
 	data->Z = Calloc(double, n*(num_eigen+1));
diff --git a/src/msvmmaj_pred.c b/src/msvmmaj_pred.c
index b551f12..6e2b3b9 100644
--- a/src/msvmmaj_pred.c
+++ b/src/msvmmaj_pred.c
@@ -196,7 +196,6 @@ void msvmmaj_predict_labels_kernel(struct MajData *data_test,
 			}
 		}
 		predy[i] = label;
-		//printf("predy[%li] = %li\n", i, label);
 	}
 
 	free(ZV);
diff --git a/src/msvmmaj_train.c b/src/msvmmaj_train.c
index 0f42ff6..2333544 100644
--- a/src/msvmmaj_train.c
+++ b/src/msvmmaj_train.c
@@ -84,7 +84,7 @@ void msvmmaj_optimize(struct MajModel *model, struct MajData *data)
 		msvmmaj_get_update(model, data, B, ZAZ, ZAZV, ZAZVT);
 		if (it > 50)
 			msvmmaj_step_doubling(model);
-
+		
 		Lbar = L;
 		L = msvmmaj_get_loss(model, data, ZV);
 
@@ -217,6 +217,10 @@ double msvmmaj_get_loss(struct MajModel *model, struct MajData *data,
  * @f]
  * solving this system is done through dposv().
  *
+ * @todo
+ * Consider allocating IPIV and WORK at a higher level, they probably don't 
+ * change much during the iterations.
+ *
  * @param [in,out] 	model 	model to be updated
  * @param [in] 		data 	data used in model
  * @param [in] 		B 	pre-allocated matrix used for linear coefficients
@@ -408,6 +412,7 @@ void msvmmaj_get_update(struct MajModel *model, struct MajData *data, double *B,
 			0.0,
 			ZAZV, 
 			K-1);
+
 	cblas_dgemm(
 			CblasRowMajor,
 			CblasTrans,
@@ -446,6 +451,7 @@ void msvmmaj_get_update(struct MajModel *model, struct MajData *data, double *B,
 	// because we have used the upper part in the BLAS
 	// calls above in Row-major order, and Lapack uses
 	// column major order. 
+
 	status = dposv(
 			'L',
 			m+1,
@@ -489,10 +495,12 @@ void msvmmaj_get_update(struct MajModel *model, struct MajData *data, double *B,
 		if (status != 0)
 			fprintf(stderr, "Received nonzero status from "
 					"dsysv: %i\n", status);
+		free(WORK);
+		free(IPIV);
 	}
 
-	// Return to Row-major order. The matrix ZAZVT contains the 
-	// solution after the dposv/dsysv call.
+	// Return to Row-major order. The matrix ZAZVT contains the solution 
+	// after the dposv/dsysv call.
 	for (i=0; i<m+1; i++)
 		for (j=0; j<K-1; j++)
 			ZAZV[i*(K-1)+j] = ZAZVT[j*(m+1)+i];
@@ -509,7 +517,7 @@ void msvmmaj_get_update(struct MajModel *model, struct MajData *data, double *B,
 	model->V = ZAZVT;
 	ZAZVT = ptr;
 	*/
-	
+
 	for (i=0; i<m+1; i++) {
 		for (j=0; j<K-1; j++) {
 			matrix_set(
diff --git a/src/msvmmaj_train_dataset.c b/src/msvmmaj_train_dataset.c
index 336be69..29b410b 100644
--- a/src/msvmmaj_train_dataset.c
+++ b/src/msvmmaj_train_dataset.c
@@ -330,9 +330,13 @@ void consistency_repeats(struct Queue *q, long repeats, TrainType traintype)
 		task = get_next_task(nq);
 		make_model_from_task(task, model);
 
-		model->n = task->train_data->n;
-		model->m = task->train_data->m;
-		model->K = task->train_data->K;
+		if (i == 0) {
+			model->n = 0;
+			model->m = task->train_data->m;
+			model->K = task->train_data->K;
+			msvmmaj_allocate_model(model);
+			msvmmaj_seed_model_V(NULL, model);
+		}
 
 		time[i] = 0.0;
 		note("(%02li/%02li:%03li)\t", i+1, N, task->ID);
@@ -401,6 +405,8 @@ void consistency_repeats(struct Queue *q, long repeats, TrainType traintype)
 		p += 1.0;
 	}
 
+	free(nq->tasks);
+	free(nq);
 	free(task);
 	free(model);
 	free(perf);
@@ -448,13 +454,13 @@ double cross_validation(struct MajModel *model, struct MajData *data,
 	msvmmaj_make_cv_split(data->n, folds, cv_idx);
 
 	for (f=0; f<folds; f++) {
-		//printf("Fold: %li\n", f);
+		note(".");
 		msvmmaj_get_tt_split(data, train_data, test_data, cv_idx, f);
 	
 		msvmmaj_make_kernel(model, train_data);
 
 		// reallocate the model if necessary for the new train split
-		msvmmaj_reallocate_model(model, train_data->n);
+		msvmmaj_reallocate_model(model, train_data->n, train_data->m);
 
 		msvmmaj_initialize_weights(train_data, model);
 
@@ -466,7 +472,7 @@ double cross_validation(struct MajModel *model, struct MajData *data,
 
 		// calculate prediction performance on test set
 		predy = Calloc(long, test_data->n);
-		msvmmaj_predict_labels(test_data, model, predy);
+		msvmmaj_predict_labels(test_data, train_data, model, predy);
 		performance = msvmmaj_prediction_perf(test_data, predy);
 		total_perf += performance * test_data->n;
 
@@ -540,44 +546,72 @@ void start_training_cv(struct Queue *q)
 	msvmmaj_free_model(model);
 }
 
-void msvmmaj_reallocate_model(struct MajModel *model, long n)
+void msvmmaj_reallocate_model(struct MajModel *model, long n, long m)
 {
 	long K = model->K;
-
-	model->UU = (double *) realloc(model->UU, n*K*(K-1)*sizeof(double));
-	if (model->UU == NULL) {
-		fprintf(stderr, "Failed to reallocate UU\n");
-		exit(1);
-	}
-
-	model->Q = (double *) realloc(model->Q, n*K*sizeof(double));
-	if (model->Q == NULL) {
-		fprintf(stderr, "Failed to reallocate Q\n");
-		exit(1);
-	}
-
-	model->H = (double *) realloc(model->H, n*K*sizeof(double));
-	if (model->H == NULL) {
-		fprintf(stderr, "Failed to reallocate H\n");
-		exit(1);
+	
+	if (model->n == n && model->m == m)
+		return;
+	if (model->n != n) {
+		model->UU = (double *) realloc(model->UU,
+			       	n*K*(K-1)*sizeof(double));
+		if (model->UU == NULL) {
+			fprintf(stderr, "Failed to reallocate UU\n");
+			exit(1);
+		}
+	
+		model->Q = (double *) realloc(model->Q, n*K*sizeof(double));
+		if (model->Q == NULL) {
+			fprintf(stderr, "Failed to reallocate Q\n");
+			exit(1);
+		}
+	
+		model->H = (double *) realloc(model->H, n*K*sizeof(double));
+		if (model->H == NULL) {
+			fprintf(stderr, "Failed to reallocate H\n");
+			exit(1);
+		}
+	
+		model->R = (double *) realloc(model->R, n*K*sizeof(double));
+		if (model->R == NULL) {
+			fprintf(stderr, "Failed to reallocate R\n");
+			exit(1);
+		}
+	
+		model->rho = (double *) realloc(model->rho, n*sizeof(double));
+		if (model->rho == NULL) {
+			fprintf(stderr, "Failed to reallocte rho\n");
+			exit(1);
+		}
+	
+		model->n = n;
 	}
+	if (model->m != m) {
+		model->W = (double *) realloc(model->W,
+			       	m*(K-1)*sizeof(double));
+		if (model->W == NULL) {
+			fprintf(stderr, "Failed to reallocate W\n");
+			exit(1);
+		}
 
-	model->R = (double *) realloc(model->R, n*K*sizeof(double));
-	if (model->R == NULL) {
-		fprintf(stderr, "Failed to reallocate R\n");
-		exit(1);
-	}
+		model->V = (double *) realloc(model->V,
+			       	(m+1)*(K-1)*sizeof(double));
+		if (model->V == NULL) {
+			fprintf(stderr, "Failed to reallocate V\n");
+			exit(1);
+		}
 
-	model->rho = (double *) realloc(model->rho, n*sizeof(double));
-	if (model->rho == NULL) {
-		fprintf(stderr, "Failed to reallocte rho\n");
-		exit(1);
+		model->Vbar = (double *) realloc(model->Vbar,
+				(m+1)*(K-1)*sizeof(double));
+		if (model->Vbar == NULL) {
+			fprintf(stderr, "Failed to reallocate Vbar\n");
+			exit(1);
+		}
+		
+		model->m = m;
 	}
-
-	model->n = n;
 }
-
-
+	
 /**
  * @brief Run the grid search for a train/test dataset
  *
@@ -677,13 +711,24 @@ void start_training_tt(struct Queue *q)
  */
 void free_queue(struct Queue *q)
 {
+	printf("\there 0\n");
 	long i;
+	printf("\there 1\n");
 	for (i=0; i<q->N; i++) {
+		printf("\there 2\n");
+		fflush(stdout);
 		free(q->tasks[i]->kernelparam);
+		printf("\there 3\n");
+		fflush(stdout);
 		free(q->tasks[i]);
+		printf("\there 4\n");
+		fflush(stdout);
 	}
+	printf("\there 5\n");
 	free(q->tasks);
+	printf("\there 6\n");
 	free(q);
+	printf("\there 7\n");
 }
 
 /**
@@ -775,7 +820,7 @@ void print_progress_string(struct Task *task, long N)
 				task->kernelparam[1]);
 	if (task->kerneltype == K_POLY || task->kerneltype == K_SIGMOID ||
 			task->kerneltype == K_RBF)
-		sprintf(buffer + strlen(buffer), "g = %2.2f\t",
+		sprintf(buffer + strlen(buffer), "g = %3.3f\t",
 				task->kernelparam[0]);
 	sprintf(buffer + strlen(buffer), "eps = %g\tw = %i\tk = %2.2f\t"
 			"l = %f\tp = %2.2f\t", task->epsilon,
diff --git a/src/trainMSVMMajdataset.c b/src/trainMSVMMajdataset.c
index 1929584..35f4611 100644
--- a/src/trainMSVMMajdataset.c
+++ b/src/trainMSVMMajdataset.c
@@ -112,10 +112,15 @@ int main(int argc, char **argv)
 		consistency_repeats(q, training->repeats, training->traintype);
 	}
 
+	printf("here 0\n");
 	free_queue(q);
+	printf("here 1\n");
 	free(training);
+	printf("here 2\n");
 	msvmmaj_free_data(train_data);
+	printf("here 3\n");
 	msvmmaj_free_data(test_data);
+	printf("here 4\n");
 
 	note("Done.\n");
 	return 0;