def test_sinc_n_components(): rng = np.random.RandomState(42) X = rng.uniform(-1, 1, [100, 1]) Y_true = np.sign(np.cos(4. * np.pi * X)) ka_gt = kanalysis(X, Y_true, n_components=len(X), quantiles=QUANTILES) assert abs(ka_gt[0] - 1) < EPSILON auc_gt = ka_gt.mean() assert abs(ka_gt[1] - 0.97858424670806643) < EPSILON assert abs(ka_gt[6] - 0.77621800324925161) < EPSILON assert abs(auc_gt - 0.26515079212133352) < EPSILON n_components = 10 ka_gv = kanalysis(X, Y_true, n_components=n_components, quantiles=QUANTILES) assert abs(ka_gv[0] - 1) < EPSILON assert (abs(ka_gv - ka_gt[:n_components + 1]) < EPSILON).all()
def test_sinc_n_components(): rng = np.random.RandomState(42) X = rng.uniform(-1, 1, [100, 1]) Y_true = np.sign(np.cos(4. * np.pi * X)) ka_gt = kanalysis(X, Y_true, n_components=len(X), quantiles=QUANTILES) assert abs(ka_gt[0] - 1) < EPSILON auc_gt = ka_gt.mean() assert abs(ka_gt[1] - 0.97858424670806643) < EPSILON assert abs(ka_gt[6] - 0.77621800324925161) < EPSILON assert abs(auc_gt - 0.26515079212133352) < EPSILON n_components = 10 ka_gv = kanalysis(X, Y_true, n_components=n_components, quantiles=QUANTILES) assert abs(ka_gv[0] - 1) < EPSILON assert (abs(ka_gv - ka_gt[:n_components + 1]) < EPSILON).all()
def test_sinc(): rng = np.random.RandomState(42) X = rng.uniform(-1, 1, [100, 1]) Y_true = np.sign(np.cos(4. * np.pi * X)) ka = kanalysis(X, Y_true, quantiles=QUANTILES) assert abs(ka[0] - 1) < EPSILON auc = ka.mean() assert abs(ka[1] - 0.97858424670806643) < EPSILON assert abs(ka[6] - 0.77621800324925161) < EPSILON assert abs(auc - 0.26515079212133352) < EPSILON X2 = np.cos(4*np.pi*X) ka2 = kanalysis(X2, Y_true, quantiles=QUANTILES) assert abs(ka2[0] - 1) < EPSILON auc2 = ka2.mean() assert abs(ka2[2] - 0.18664740171388722) < EPSILON assert abs(ka2[8] - 0.08675467229138914) < EPSILON assert abs(auc2 - 0.054521533910044662) < EPSILON
def test_sinc(): rng = np.random.RandomState(42) X = rng.uniform(-1, 1, [100, 1]) Y_true = np.sign(np.cos(4. * np.pi * X)) ka = kanalysis(X, Y_true, quantiles=QUANTILES) assert abs(ka[0] - 1) < EPSILON auc = ka.mean() assert abs(ka[1] - 0.97858424670806643) < EPSILON assert abs(ka[6] - 0.77621800324925161) < EPSILON assert abs(auc - 0.26515079212133352) < EPSILON X2 = np.cos(4 * np.pi * X) ka2 = kanalysis(X2, Y_true, quantiles=QUANTILES) assert abs(ka2[0] - 1) < EPSILON auc2 = ka2.mean() assert abs(ka2[2] - 0.18664740171388722) < EPSILON assert abs(ka2[8] - 0.08675467229138914) < EPSILON assert abs(auc2 - 0.054521533910044662) < EPSILON
def test_simple2d(): rng = np.random.RandomState(42) X = rng.randn(8, 4) Y = rng.randn(*X.shape) gv = kanalysis(X, Y) assert abs(gv[0] - 1) < EPSILON gt = np.array([ 1.00000000e+00, 8.36576988e-01, 7.62676568e-01, 6.14490305e-01, 3.61717470e-01, 2.32448139e-01, 1.12094204e-01, 9.04488924e-31, 7.08820460e-31, ]) assert (abs(gv - gt) < EPSILON).all()
def test_simple1d(): rng = np.random.RandomState(42) X = rng.randn(8, 4) Y = rng.randn(8) gv = kanalysis(X, Y) assert abs(gv[0] - 1) < EPSILON gt = np.array([ 1.00000000e+00, 9.71468057e-01, 5.59939220e-01, 5.20056617e-01, 5.04160095e-01, 3.38182408e-01, 3.12550170e-01, 1.50530684e-30, 1.03692068e-30, ]) assert (abs(gv - gt) < EPSILON).all()
def test_simple2d(): rng = np.random.RandomState(42) X = rng.randn(8, 4) Y = rng.randn(*X.shape) gv = kanalysis(X, Y) assert abs(gv[0] - 1) < EPSILON gt = np.array([ 1.00000000e+00, 8.36576988e-01, 7.62676568e-01, 6.14490305e-01, 3.61717470e-01, 2.32448139e-01, 1.12094204e-01, 9.04488924e-31, 7.08820460e-31, ]) assert (abs(gv - gt) < EPSILON).all()
def test_simple1d(): rng = np.random.RandomState(42) X = rng.randn(8, 4) Y = rng.randn(8) gv = kanalysis(X, Y) assert abs(gv[0] - 1) < EPSILON gt = np.array([ 1.00000000e+00, 9.71468057e-01, 5.59939220e-01, 5.20056617e-01, 5.04160095e-01, 3.38182408e-01, 3.12550170e-01, 1.50530684e-30, 1.03692068e-30, ]) assert (abs(gv - gt) < EPSILON).all()
def test_smoke_size(): rng = np.random.RandomState(42) X = rng.randn(10, 4) Y = rng.randn(*X.shape) gv = kanalysis(X, Y) assert gv.size == len(X) + 1
def test_smoke_size(): rng = np.random.RandomState(42) X = rng.randn(10, 4) Y = rng.randn(*X.shape) gv = kanalysis(X, Y) assert gv.size == len(X) + 1