restart using git

1 files changed, 111 insertions, 0 deletions

diff --git a/src/msvmmaj_pred.c b/src/msvmmaj_pred.c
new file mode 100644
index 0000000..5f1b1ae
--- /dev/null
+++ b/src/msvmmaj_pred.c
@@ -0,0 +1,111 @@
+/**
+ * @file msvmmaj_pred.c
+ * @author Gertjan van den Burg (burg@ese.eur.nl)
+ * @date August 9, 2013
+ * @brief Main functions for predicting class labels..
+ *
+ */
+
+#include <cblas.h>
+
+#include "libMSVMMaj.h"
+#include "MSVMMaj.h"
+#include "matrix.h"
+#include "msvmmaj_pred.h"
+
+/**
+ * @name predict_labels
+ * @brief Predict class labels of data given and output in predy
+ *
+ * The labels are predicted by mapping each instance in data to the 
+ * simplex space using the matrix V in the given model. Next, for each
+ * instance the nearest simplex vertex is determined using an Euclidean
+ * norm. The nearest simplex vertex determines the predicted class label,
+ * which is recorded in predy
+ *
+ * @param [in] 		data 		data to predict labels for
+ * @param [in] 		model 		model with optimized V
+ * @param [out] 	predy 		pre-allocated vector to record predictions in
+ */
+void msvmmaj_predict_labels(struct MajData *data, struct MajModel *model, long *predy)
+{
+	long i, j, k, label;
+	double norm, min_dist;
+
+	long n = data->n; // note that model->n is the size of the training sample.
+	long m = data->m;
+	long K = model->K; //data->K does not necessarily equal the original K.
+
+	double *S = Calloc(double, K-1);
+	double *ZV = Calloc(double, n*(K-1));
+	double *U = Calloc(double, K*(K-1));
+
+	// Get the simplex matrix
+	msvmmaj_simplex_gen(K, U);
+
+	// Generate the simplex-space vectors
+	cblas_dgemm(
+			CblasRowMajor,
+			CblasNoTrans,
+			CblasNoTrans,
+			n,
+			K-1,
+			m+1,
+			1.0,
+			data->Z,
+			m+1,
+			model->V,
+			K-1,
+			0.0,
+			ZV,
+			K-1);
+
+	// Calculate the distance to each of the vertices of the simplex.
+	// The closest vertex defines the class label.
+	for (i=0; i<n; i++) {
+		label = 0;
+		min_dist = 1000000000.0;
+		for (j=0; j<K; j++) {
+			for (k=0; k<K-1; k++) {
+				S[k] = matrix_get(ZV, K-1, i, k) - matrix_get(U, K-1, j, k);
+			}
+			norm = cblas_dnrm2(K, S, 1);
+			if (norm < min_dist) {
+				label = j+1; // labels start counting from 1
+				min_dist = norm;
+			}
+		}
+		predy[i] = label;
+	}
+
+	free(ZV);
+	free(U);
+	free(S);
+}
+
+/**
+ * @name msvmmaj_prediction_perf
+ * @brief Calculate the predictive performance (percentage correct)
+ *
+ * The predictive performance is calculated by simply counting the number
+ * of correctly classified samples and dividing by the total number of 
+ * samples, multiplying by 100.
+ *
+ * @param [in] 		data 		the dataset with known labels
+ * @param [in] 		predy 		the predicted class labels
+ *
+ * @returns percentage correctly classified.
+ */
+double msvmmaj_prediction_perf(struct MajData *data, long *predy)
+{
+	long i, correct = 0;
+	double performance;
+
+	for (i=0; i<data->n; i++)
+		if (data->y[i] == predy[i])
+			correct++;
+
+	performance = ((double) correct)/((double) data->n)* 100.0;
+
+	return performance;
+}