sklearn SVM: actually pass gamma to SVC

Author: bibikar

sklearn/svm.py

@@ -73,7 +73,7 @@ def get_optimal_cache_size(n_features, dtype=np.double, max_cache=64):
 # Create our C-SVM classifier
 clf = SVC(C=params.C, kernel=params.kernel, max_iter=params.maxiter,
           cache_size=params.cache_size_mb, tol=params.tol,
-          shrinking=params.shrinking, gamma='auto')
+          shrinking=params.shrinking, gamma=params.gamma)
 
 columns = ('batch', 'arch', 'prefix', 'function', 'threads', 'dtype', 'size',
            'kernel', 'cache_size_mb', 'C', 'sv_len', 'n_classes', 'accuracy',