019ReviewmidPython / scikit-learn真实来源改编 scikit-learn
scikit-learn SVM：处理空 support vectors 的 sparse fit
审查一个 scikit-learn SVM PR：作者在 n_SV 为 0 时走专门分支，避免稀疏 dual_coef_ 构造报错。
@@ -287,11 +287,17 @@ class BaseLibSVM(BaseEstimator, metaclass=ABCMeta): n_class = getattr(self, "_n_class", 1) n_SV = self.support_vectors_.shape[0] dual_coef_indices = np.tile(np.arange(n_SV), n_class)- dual_coef_indptr = np.arange(- 0,- dual_coef_indices.size + 1,- dual_coef_indices.size / n_class,- )- self.dual_coef_ = sp.csr_matrix(- (dual_coef_data, dual_coef_indices, dual_coef_indptr),- (n_class, n_SV),- )+ if not n_SV:+ self.dual_coef_ = sp.csr_matrix([])+ else:+ dual_coef_indptr = np.arange(+ 0,+ dual_coef_indices.size + 1,+ dual_coef_indices.size / n_class,+ )+ self.dual_coef_ = sp.csr_matrix(+ (dual_coef_data, dual_coef_indices, dual_coef_indptr),+ (n_class, n_SV),+ ) return self@@ -1005,6 +1005,17 @@ def test_svm_regression(): assert_allclose(pred, y, atol=0.5) +def test_sparse_fit_support_vectors_empty():+ # Regression test for gh-14894: a sparse SVR whose fit produces no+ # support vectors used to raise ZeroDivisionError while building+ # dual_coef_indptr (step = 0). It must fit cleanly instead.+ X_train = sparse.csr_matrix([[0, 1, 0, 0],+ [0, 0, 0, 1],+ [0, 0, 1, 0],+ [0, 0, 0, 1]])+ y_train = np.array([0.04, 0.04, 0.10, 0.16])+ model = svm.SVR(kernel='linear')+ model.fit(X_train, y_train)+ assert not model.support_vectors_.data.size+ assert not model.dual_coef_.data.size++ def test_linearsvc_parameters(): # Test possible parameter combinations in LinearSVC