Fix svm using for sklearn and cuml (#28)

Author: PetrovKP

cuml/svm.py

@@ -36,26 +36,26 @@ def get_optimal_cache_size(n_rows, dtype=np.double, max_cache=64):
 
 parser = argparse.ArgumentParser(description='cuML SVM benchmark')
 
-parser.add_argument('-C', dest='C', type=float, default=0.01,
-                    help='SVM slack parameter')
+parser.add_argument('-C', dest='C', type=float, default=1.0,
+                    help='SVM regularization parameter')
 parser.add_argument('--kernel', choices=('linear', 'rbf'),
                     default='linear', help='SVM kernel function')
-parser.add_argument('--gamma', type=float, default=None,
-                    help='Parameter for kernel="rbf"')
-parser.add_argument('--maxiter', type=int, default=2000,
+parser.add_argument('--maxiter', type=int, default=-1,
                     help='Maximum iterations for the iterative solver. '
                          '-1 means no limit.')
-parser.add_argument('--max-cache-size', type=int, default=64,
+parser.add_argument('--gamma', type=float, default=None,
+                    help='Parameter for kernel="rbf"')
+parser.add_argument('--max-cache-size', type=int, default=8,
                     help='Maximum cache size, in gigabytes, for SVM.')
-parser.add_argument('--tol', type=float, default=1e-16,
+parser.add_argument('--tol', type=float, default=1e-3,
                     help='Tolerance passed to sklearn.svm.SVC')
 params = parse_args(parser)
 
 # Load data
 X_train, X_test, y_train, y_test = load_data(params)
 
 if params.gamma is None:
-    params.gamma = 'auto'
+    params.gamma = 1.0 / X_train.shape[1]
 
 cache_size_bytes = get_optimal_cache_size(X_train.shape[0],
                                           max_cache=params.max_cache_size)

daal4py/svm.py

@@ -269,6 +269,7 @@ def test_predict(X, training_result, params):
     else:
         prdct = multi_class_classifier_prediction(
                 nClasses=params.n_classes,
+                method='thunder',
                 fptype=fptype,
                 maxIterations=params.maxiter,
                 accuracyThreshold=params.tol,
@@ -290,20 +291,19 @@ def test_predict(X, training_result, params):
 def main():
     parser = argparse.ArgumentParser(description='daal4py SVC benchmark with '
                                                  'linear kernel')
-    parser.add_argument('-C', dest='C', type=float, default=0.01,
-                        help='SVM slack parameter')
+    parser.add_argument('-C', dest='C', type=float, default=1.0,
+                        help='SVM regularization parameter')
     parser.add_argument('--kernel', choices=('linear', 'rbf'),
                         default='linear', help='SVM kernel function')
     parser.add_argument('--gamma', type=float, default=None,
                         help='Parameter for kernel="rbf"')
-    parser.add_argument('--maxiter', type=int, default=2000,
-                        help='Maximum iterations for the iterative solver. '
-                             '-1 means no limit.')
-    parser.add_argument('--max-cache-size', type=int, default=64,
+    parser.add_argument('--maxiter', type=int, default=100000,
+                        help='Maximum iterations for the iterative solver. ')
+    parser.add_argument('--max-cache-size', type=int, default=8,
                         help='Maximum cache size, in gigabytes, for SVM.')
     parser.add_argument('--tau', type=float, default=1e-12,
                         help='Tau parameter for working set selection scheme')
-    parser.add_argument('--tol', type=float, default=1e-16,
+    parser.add_argument('--tol', type=float, default=1e-3,
                         help='Tolerance')
     parser.add_argument('--no-shrinking', action='store_false', default=True,
                         dest='shrinking',

sklearn/svm.py

@@ -37,18 +37,18 @@ def get_optimal_cache_size(n_rows, dtype=np.double, max_cache=64):
 
 parser = argparse.ArgumentParser(description='scikit-learn SVM benchmark')
 
-parser.add_argument('-C', dest='C', type=float, default=0.01,
-                    help='SVM slack parameter')
+parser.add_argument('-C', dest='C', type=float, default=1.0,
+                    help='SVM regularization parameter')
 parser.add_argument('--kernel', choices=('linear', 'rbf'),
                     default='linear', help='SVM kernel function')
 parser.add_argument('--gamma', type=float, default=None,
                     help='Parameter for kernel="rbf"')
-parser.add_argument('--maxiter', type=int, default=2000,
+parser.add_argument('--maxiter', type=int, default=-1,
                     help='Maximum iterations for the iterative solver. '
                          '-1 means no limit.')
-parser.add_argument('--max-cache-size', type=int, default=64,
+parser.add_argument('--max-cache-size', type=int, default=8,
                     help='Maximum cache size, in gigabytes, for SVM.')
-parser.add_argument('--tol', type=float, default=1e-16,
+parser.add_argument('--tol', type=float, default=1e-3,
                     help='Tolerance passed to sklearn.svm.SVC')
 parser.add_argument('--no-shrinking', action='store_false', default=True,
                     dest='shrinking', help="Don't use shrinking heuristic")
@@ -58,7 +58,7 @@ def get_optimal_cache_size(n_rows, dtype=np.double, max_cache=64):
 X_train, X_test, y_train, y_test = load_data(params)
 
 if params.gamma is None:
-    params.gamma = 'auto'
+    params.gamma = 1.0 / X_train.shape[1]
 
 cache_size_bytes = get_optimal_cache_size(X_train.shape[0],
                                           max_cache=params.max_cache_size)