some debugs

cnn.py

@@ -181,7 +181,10 @@ def run_kfold(X_train, X_test, y_train, y_test, params):
 		print("Area under the curve:", auc)
 
 		# confusion matrix
-		cm+=confusion_matrix(y_test[i].argmax(axis=1), y_pred.argmax(axis=1))
+		if i == 0:
+			cm = confusion_matrix(y_test[i].argmax(axis=1), y_pred.argmax(axis=1))
+		else:
+			cm+=confusion_matrix(y_test[i].argmax(axis=1), y_pred.argmax(axis=1))
 
 		#pr curve (contains 4 pr curves: one for each class)
 		recall, precision, average_prec = pr(N_CLASSES, y_test[i], y_pred)