scipy 1.16.2__cp313-cp313t-win_arm64.whl

scipy/linalg/tests/test_matmul_toeplitz.py

@@ -0,0 +1,136 @@
+"""Test functions for linalg.matmul_toeplitz function
+"""
+
+import numpy as np
+from scipy.linalg import toeplitz, matmul_toeplitz
+
+from pytest import raises as assert_raises
+from numpy.testing import assert_allclose
+
+
+class TestMatmulToeplitz:
+
+    def setup_method(self):
+        self.rng = np.random.RandomState(42)
+        self.tolerance = 1.5e-13
+
+    def test_real(self):
+        cases = []
+
+        n = 1
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n)
+        x = self.rng.normal(size=(n, 1))
+        cases.append((x, c, r, False))
+
+        n = 2
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n)
+        x = self.rng.normal(size=(n, 1))
+        cases.append((x, c, r, False))
+
+        n = 101
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n)
+        x = self.rng.normal(size=(n, 1))
+        cases.append((x, c, r, True))
+
+        n = 1000
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n)
+        x = self.rng.normal(size=(n, 1))
+        cases.append((x, c, r, False))
+
+        n = 100
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n)
+        x = self.rng.normal(size=(n, self.rng.randint(1, 10)))
+        cases.append((x, c, r, False))
+
+        n = 100
+        c = self.rng.normal(size=(n, 1))
+        r = self.rng.normal(size=(n, 1))
+        x = self.rng.normal(size=(n, self.rng.randint(1, 10)))
+        cases.append((x, c, r, True))
+
+        n = 100
+        c = self.rng.normal(size=(n, 1))
+        r = None
+        x = self.rng.normal(size=(n, self.rng.randint(1, 10)))
+        cases.append((x, c, r, True, -1))
+
+        n = 100
+        c = self.rng.normal(size=(n, 1))
+        r = None
+        x = self.rng.normal(size=n)
+        cases.append((x, c, r, False))
+
+        n = 101
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n-27)
+        x = self.rng.normal(size=(n-27, 1))
+        cases.append((x, c, r, True))
+
+        n = 100
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n//4)
+        x = self.rng.normal(size=(n//4, self.rng.randint(1, 10)))
+        cases.append((x, c, r, True))
+
+        [self.do(*i) for i in cases]
+
+    def test_complex(self):
+        n = 127
+        c = self.rng.normal(size=(n, 1)) + self.rng.normal(size=(n, 1))*1j
+        r = self.rng.normal(size=(n, 1)) + self.rng.normal(size=(n, 1))*1j
+        x = self.rng.normal(size=(n, 3)) + self.rng.normal(size=(n, 3))*1j
+        self.do(x, c, r, False)
+
+        n = 100
+        c = self.rng.normal(size=(n, 1)) + self.rng.normal(size=(n, 1))*1j
+        r = self.rng.normal(size=(n//2, 1)) +\
+            self.rng.normal(size=(n//2, 1))*1j
+        x = self.rng.normal(size=(n//2, 3)) +\
+            self.rng.normal(size=(n//2, 3))*1j
+        self.do(x, c, r, False)
+
+    def test_empty(self):
+        c = []
+        r = []
+        x = []
+        self.do(x, c, r, False)
+
+        x = np.empty((0, 0))
+        self.do(x, c, r, False)
+
+    def test_exceptions(self):
+
+        n = 100
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=2*n)
+        x = self.rng.normal(size=n)
+        assert_raises(ValueError, matmul_toeplitz, (c, r), x, True)
+
+        n = 100
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n)
+        x = self.rng.normal(size=n-1)
+        assert_raises(ValueError, matmul_toeplitz, (c, r), x, True)
+
+        n = 100
+        c = self.rng.normal(size=n)
+        r = self.rng.normal(size=n//2)
+        x = self.rng.normal(size=n//2-1)
+        assert_raises(ValueError, matmul_toeplitz, (c, r), x, True)
+
+    # For toeplitz matrices, matmul_toeplitz() should be equivalent to @.
+    def do(self, x, c, r=None, check_finite=False, workers=None):
+        c = np.ravel(c)
+        if r is None:
+            actual = matmul_toeplitz(c, x, check_finite, workers)
+        else:
+            r = np.ravel(r)
+            actual = matmul_toeplitz((c, r), x, check_finite)
+        desired = toeplitz(c, r) @ x
+        assert_allclose(actual, desired,
+            rtol=self.tolerance, atol=self.tolerance)

scipy/linalg/tests/test_procrustes.py

@@ -0,0 +1,214 @@
+from itertools import product, permutations
+
+import numpy as np
+import pytest
+from numpy.testing import assert_array_less, assert_allclose
+from pytest import raises as assert_raises
+
+from scipy.linalg import inv, eigh, norm, svd
+from scipy.linalg import orthogonal_procrustes
+from scipy.sparse._sputils import matrix
+
+
+def test_orthogonal_procrustes_ndim_too_small():
+    rng = np.random.RandomState(1234)
+    A = rng.randn(3)
+    B = rng.randn(3)
+    assert_raises(ValueError, orthogonal_procrustes, A, B)
+
+
+def test_orthogonal_procrustes_shape_mismatch():
+    rng = np.random.RandomState(1234)
+    shapes = ((3, 3), (3, 4), (4, 3), (4, 4))
+    for a, b in permutations(shapes, 2):
+        A = rng.randn(*a)
+        B = rng.randn(*b)
+        assert_raises(ValueError, orthogonal_procrustes, A, B)
+
+
+def test_orthogonal_procrustes_checkfinite_exception():
+    rng = np.random.RandomState(1234)
+    m, n = 2, 3
+    A_good = rng.randn(m, n)
+    B_good = rng.randn(m, n)
+    for bad_value in np.inf, -np.inf, np.nan:
+        A_bad = A_good.copy()
+        A_bad[1, 2] = bad_value
+        B_bad = B_good.copy()
+        B_bad[1, 2] = bad_value
+        for A, B in ((A_good, B_bad), (A_bad, B_good), (A_bad, B_bad)):
+            assert_raises(ValueError, orthogonal_procrustes, A, B)
+
+
+def test_orthogonal_procrustes_scale_invariance():
+    rng = np.random.RandomState(1234)
+    m, n = 4, 3
+    for i in range(3):
+        A_orig = rng.randn(m, n)
+        B_orig = rng.randn(m, n)
+        R_orig, s = orthogonal_procrustes(A_orig, B_orig)
+        for A_scale in np.square(rng.randn(3)):
+            for B_scale in np.square(rng.randn(3)):
+                R, s = orthogonal_procrustes(A_orig * A_scale, B_orig * B_scale)
+                assert_allclose(R, R_orig)
+
+
+def test_orthogonal_procrustes_array_conversion():
+    rng = np.random.RandomState(1234)
+    for m, n in ((6, 4), (4, 4), (4, 6)):
+        A_arr = rng.randn(m, n)
+        B_arr = rng.randn(m, n)
+        As = (A_arr, A_arr.tolist(), matrix(A_arr))
+        Bs = (B_arr, B_arr.tolist(), matrix(B_arr))
+        R_arr, s = orthogonal_procrustes(A_arr, B_arr)
+        AR_arr = A_arr.dot(R_arr)
+        for A, B in product(As, Bs):
+            R, s = orthogonal_procrustes(A, B)
+            AR = A_arr.dot(R)
+            assert_allclose(AR, AR_arr)
+
+
+def test_orthogonal_procrustes():
+    rng = np.random.RandomState(1234)
+    for m, n in ((6, 4), (4, 4), (4, 6)):
+        # Sample a random target matrix.
+        B = rng.randn(m, n)
+        # Sample a random orthogonal matrix
+        # by computing eigh of a sampled symmetric matrix.
+        X = rng.randn(n, n)
+        w, V = eigh(X.T + X)
+        assert_allclose(inv(V), V.T)
+        # Compute a matrix with a known orthogonal transformation that gives B.
+        A = np.dot(B, V.T)
+        # Check that an orthogonal transformation from A to B can be recovered.
+        R, s = orthogonal_procrustes(A, B)
+        assert_allclose(inv(R), R.T)
+        assert_allclose(A.dot(R), B)
+        # Create a perturbed input matrix.
+        A_perturbed = A + 1e-2 * rng.randn(m, n)
+        # Check that the orthogonal procrustes function can find an orthogonal
+        # transformation that is better than the orthogonal transformation
+        # computed from the original input matrix.
+        R_prime, s = orthogonal_procrustes(A_perturbed, B)
+        assert_allclose(inv(R_prime), R_prime.T)
+        # Compute the naive and optimal transformations of the perturbed input.
+        naive_approx = A_perturbed.dot(R)
+        optim_approx = A_perturbed.dot(R_prime)
+        # Compute the Frobenius norm errors of the matrix approximations.
+        naive_approx_error = norm(naive_approx - B, ord='fro')
+        optim_approx_error = norm(optim_approx - B, ord='fro')
+        # Check that the orthogonal Procrustes approximation is better.
+        assert_array_less(optim_approx_error, naive_approx_error)
+
+
+def _centered(A):
+    mu = A.mean(axis=0)
+    return A - mu, mu
+
+
+def test_orthogonal_procrustes_exact_example():
+    # Check a small application.
+    # It uses translation, scaling, reflection, and rotation.
+    #
+    #         |
+    #   a  b  |
+    #         |
+    #   d  c  |        w
+    #         |
+    # --------+--- x ----- z ---
+    #         |
+    #         |        y
+    #         |
+    #
+    A_orig = np.array([[-3, 3], [-2, 3], [-2, 2], [-3, 2]], dtype=float)
+    B_orig = np.array([[3, 2], [1, 0], [3, -2], [5, 0]], dtype=float)
+    A, A_mu = _centered(A_orig)
+    B, B_mu = _centered(B_orig)
+    R, s = orthogonal_procrustes(A, B)
+    scale = s / np.square(norm(A))
+    B_approx = scale * np.dot(A, R) + B_mu
+    assert_allclose(B_approx, B_orig, atol=1e-8)
+
+
+def test_orthogonal_procrustes_stretched_example():
+    # Try again with a target with a stretched y axis.
+    A_orig = np.array([[-3, 3], [-2, 3], [-2, 2], [-3, 2]], dtype=float)
+    B_orig = np.array([[3, 40], [1, 0], [3, -40], [5, 0]], dtype=float)
+    A, A_mu = _centered(A_orig)
+    B, B_mu = _centered(B_orig)
+    R, s = orthogonal_procrustes(A, B)
+    scale = s / np.square(norm(A))
+    B_approx = scale * np.dot(A, R) + B_mu
+    expected = np.array([[3, 21], [-18, 0], [3, -21], [24, 0]], dtype=float)
+    assert_allclose(B_approx, expected, atol=1e-8)
+    # Check disparity symmetry.
+    expected_disparity = 0.4501246882793018
+    AB_disparity = np.square(norm(B_approx - B_orig) / norm(B))
+    assert_allclose(AB_disparity, expected_disparity)
+    R, s = orthogonal_procrustes(B, A)
+    scale = s / np.square(norm(B))
+    A_approx = scale * np.dot(B, R) + A_mu
+    BA_disparity = np.square(norm(A_approx - A_orig) / norm(A))
+    assert_allclose(BA_disparity, expected_disparity)
+
+
+def test_orthogonal_procrustes_skbio_example():
+    # This transformation is also exact.
+    # It uses translation, scaling, and reflection.
+    #
+    #   |
+    #   | a
+    #   | b
+    #   | c d
+    # --+---------
+    #   |
+    #   |       w
+    #   |
+    #   |       x
+    #   |
+    #   |   z   y
+    #   |
+    #
+    A_orig = np.array([[4, -2], [4, -4], [4, -6], [2, -6]], dtype=float)
+    B_orig = np.array([[1, 3], [1, 2], [1, 1], [2, 1]], dtype=float)
+    B_standardized = np.array([
+        [-0.13363062, 0.6681531],
+        [-0.13363062, 0.13363062],
+        [-0.13363062, -0.40089186],
+        [0.40089186, -0.40089186]])
+    A, A_mu = _centered(A_orig)
+    B, B_mu = _centered(B_orig)
+    R, s = orthogonal_procrustes(A, B)
+    scale = s / np.square(norm(A))
+    B_approx = scale * np.dot(A, R) + B_mu
+    assert_allclose(B_approx, B_orig)
+    assert_allclose(B / norm(B), B_standardized)
+
+
+def test_empty():
+    a = np.empty((0, 0))
+    r, s = orthogonal_procrustes(a, a)
+    assert_allclose(r, np.empty((0, 0)))
+
+    a = np.empty((0, 3))
+    r, s = orthogonal_procrustes(a, a)
+    assert_allclose(r, np.identity(3))
+
+
+@pytest.mark.parametrize('shape', [(4, 5), (5, 5), (5, 4)])
+def test_unitary(shape):
+    # gh-12071 added support for unitary matrices; check that it
+    # works as intended.
+    m, n = shape
+    rng = np.random.default_rng(589234981235)
+    A = rng.random(shape) + rng.random(shape) * 1j
+    Q = rng.random((n, n)) + rng.random((n, n)) * 1j
+    Q, _ = np.linalg.qr(Q)
+    B = A @ Q
+    R, scale = orthogonal_procrustes(A, B)
+    assert_allclose(R @ R.conj().T, np.eye(n), atol=1e-14)
+    assert_allclose(A @ Q, B)
+    if shape != (4, 5):  # solution is unique
+        assert_allclose(R, Q)
+    _, s, _ = svd(A.conj().T @ B)
+    assert_allclose(scale, np.sum(s))

scipy/linalg/tests/test_sketches.py

@@ -0,0 +1,118 @@
+"""Tests for _sketches.py."""
+
+import numpy as np
+from numpy.testing import assert_, assert_equal
+from scipy.linalg import clarkson_woodruff_transform
+from scipy.linalg._sketches import cwt_matrix
+from scipy.sparse import issparse, rand
+from scipy.sparse.linalg import norm
+
+
+class TestClarksonWoodruffTransform:
+    """
+    Testing the Clarkson Woodruff Transform
+    """
+    # set seed for generating test matrices
+    rng = np.random.default_rng(1179103485)
+
+    # Test matrix parameters
+    n_rows = 2000
+    n_cols = 100
+    density = 0.1
+
+    # Sketch matrix dimensions
+    n_sketch_rows = 200
+
+    # Seeds to test with
+    seeds = [1755490010, 934377150, 1391612830, 1752708722, 2008891431,
+             1302443994, 1521083269, 1501189312, 1126232505, 1533465685]
+
+    A_dense = rng.random((n_rows, n_cols))
+    A_csc = rand(
+        n_rows, n_cols, density=density, format='csc', random_state=rng,
+    )
+    A_csr = rand(
+        n_rows, n_cols, density=density, format='csr', random_state=rng,
+    )
+    A_coo = rand(
+        n_rows, n_cols, density=density, format='coo', random_state=rng,
+    )
+
+    # Collect the test matrices
+    test_matrices = [
+        A_dense, A_csc, A_csr, A_coo,
+    ]
+
+    # Test vector with norm ~1
+    x = rng.random((n_rows, 1)) / np.sqrt(n_rows)
+
+    def test_sketch_dimensions(self):
+        for A in self.test_matrices:
+            for seed in self.seeds:
+                # seed to ensure backwards compatibility post SPEC7
+                sketch = clarkson_woodruff_transform(
+                    A, self.n_sketch_rows, seed=seed
+                )
+                assert_(sketch.shape == (self.n_sketch_rows, self.n_cols))
+
+    def test_seed_returns_identical_transform_matrix(self):
+        for seed in self.seeds:
+            S1 = cwt_matrix(
+                self.n_sketch_rows, self.n_rows, rng=seed
+            ).toarray()
+            S2 = cwt_matrix(
+                self.n_sketch_rows, self.n_rows, rng=seed
+            ).toarray()
+            assert_equal(S1, S2)
+
+    def test_seed_returns_identically(self):
+        for A in self.test_matrices:
+            for seed in self.seeds:
+                sketch1 = clarkson_woodruff_transform(
+                    A, self.n_sketch_rows, rng=seed
+                )
+                sketch2 = clarkson_woodruff_transform(
+                    A, self.n_sketch_rows, rng=seed
+                )
+                if issparse(sketch1):
+                    sketch1 = sketch1.toarray()
+                if issparse(sketch2):
+                    sketch2 = sketch2.toarray()
+                assert_equal(sketch1, sketch2)
+
+    def test_sketch_preserves_frobenius_norm(self):
+        # Given the probabilistic nature of the sketches
+        # we run the test multiple times and check that
+        # we pass all/almost all the tries.
+        n_errors = 0
+        for A in self.test_matrices:
+            if issparse(A):
+                true_norm = norm(A)
+            else:
+                true_norm = np.linalg.norm(A)
+            for seed in self.seeds:
+                sketch = clarkson_woodruff_transform(
+                    A, self.n_sketch_rows, rng=seed,
+                )
+                if issparse(sketch):
+                    sketch_norm = norm(sketch)
+                else:
+                    sketch_norm = np.linalg.norm(sketch)
+
+                if np.abs(true_norm - sketch_norm) > 0.1 * true_norm:
+                    n_errors += 1
+        assert_(n_errors == 0)
+
+    def test_sketch_preserves_vector_norm(self):
+        n_errors = 0
+        n_sketch_rows = int(np.ceil(2. / (0.01 * 0.5**2)))
+        true_norm = np.linalg.norm(self.x)
+        for seed in self.seeds:
+            sketch = clarkson_woodruff_transform(
+                self.x, n_sketch_rows, rng=seed,
+            )
+            sketch_norm = np.linalg.norm(sketch)
+
+            if np.abs(true_norm - sketch_norm) > 0.5 * true_norm:
+                n_errors += 1
+        assert_(n_errors == 0)

scipy/linalg/tests/test_solve_toeplitz.py

@@ -0,0 +1,150 @@
+"""Test functions for linalg._solve_toeplitz module
+"""
+import numpy as np
+from scipy.linalg._solve_toeplitz import levinson
+from scipy.linalg import solve, toeplitz, solve_toeplitz, matmul_toeplitz
+from numpy.testing import assert_equal, assert_allclose
+
+import pytest
+from pytest import raises as assert_raises
+
+
+def test_solve_equivalence():
+    # For toeplitz matrices, solve_toeplitz() should be equivalent to solve().
+    random = np.random.RandomState(1234)
+    for n in (1, 2, 3, 10):
+        c = random.randn(n)
+        if random.rand() < 0.5:
+            c = c + 1j * random.randn(n)
+        r = random.randn(n)
+        if random.rand() < 0.5:
+            r = r + 1j * random.randn(n)
+        y = random.randn(n)
+        if random.rand() < 0.5:
+            y = y + 1j * random.randn(n)
+
+        # Check equivalence when both the column and row are provided.
+        actual = solve_toeplitz((c,r), y)
+        desired = solve(toeplitz(c, r=r), y)
+        assert_allclose(actual, desired)
+
+        # Check equivalence when the column is provided but not the row.
+        actual = solve_toeplitz(c, b=y)
+        desired = solve(toeplitz(c), y)
+        assert_allclose(actual, desired)
+
+
+def test_multiple_rhs():
+    random = np.random.RandomState(1234)
+    c = random.randn(4)
+    r = random.randn(4)
+    for offset in [0, 1j]:
+        for yshape in ((4,), (4, 3)):
+            y = random.randn(*yshape) + offset
+            actual = solve_toeplitz((c,r), b=y)
+            desired = solve(toeplitz(c, r=r), y)
+            assert_equal(actual.shape, yshape)
+            assert_equal(desired.shape, yshape)
+            assert_allclose(actual, desired)
+
+
+def test_native_list_arguments():
+    c = [1,2,4,7]
+    r = [1,3,9,12]
+    y = [5,1,4,2]
+    actual = solve_toeplitz((c,r), y)
+    desired = solve(toeplitz(c, r=r), y)
+    assert_allclose(actual, desired)
+
+
+def test_zero_diag_error():
+    # The Levinson-Durbin implementation fails when the diagonal is zero.
+    random = np.random.RandomState(1234)
+    n = 4
+    c = random.randn(n)
+    r = random.randn(n)
+    y = random.randn(n)
+    c[0] = 0
+    assert_raises(np.linalg.LinAlgError,
+        solve_toeplitz, (c, r), b=y)
+
+
+def test_wikipedia_counterexample():
+    # The Levinson-Durbin implementation also fails in other cases.
+    # This example is from the talk page of the wikipedia article.
+    random = np.random.RandomState(1234)
+    c = [2, 2, 1]
+    y = random.randn(3)
+    assert_raises(np.linalg.LinAlgError, solve_toeplitz, c, b=y)
+
+
+def test_reflection_coeffs():
+    # check that the partial solutions are given by the reflection
+    # coefficients
+
+    random = np.random.RandomState(1234)
+    y_d = random.randn(10)
+    y_z = random.randn(10) + 1j
+    reflection_coeffs_d = [1]
+    reflection_coeffs_z = [1]
+    for i in range(2, 10):
+        reflection_coeffs_d.append(solve_toeplitz(y_d[:(i-1)], b=y_d[1:i])[-1])
+        reflection_coeffs_z.append(solve_toeplitz(y_z[:(i-1)], b=y_z[1:i])[-1])
+
+    y_d_concat = np.concatenate((y_d[-2:0:-1], y_d[:-1]))
+    y_z_concat = np.concatenate((y_z[-2:0:-1].conj(), y_z[:-1]))
+    _, ref_d = levinson(y_d_concat, b=y_d[1:])
+    _, ref_z = levinson(y_z_concat, b=y_z[1:])
+
+    assert_allclose(reflection_coeffs_d, ref_d[:-1])
+    assert_allclose(reflection_coeffs_z, ref_z[:-1])
+
+
+@pytest.mark.xfail(reason='Instability of Levinson iteration')
+def test_unstable():
+    # this is a "Gaussian Toeplitz matrix", as mentioned in Example 2 of
+    # I. Gohbert, T. Kailath and V. Olshevsky "Fast Gaussian Elimination with
+    # Partial Pivoting for Matrices with Displacement Structure"
+    # Mathematics of Computation, 64, 212 (1995), pp 1557-1576
+    # which can be unstable for levinson recursion.
+
+    # other fast toeplitz solvers such as GKO or Burg should be better.
+    random = np.random.RandomState(1234)
+    n = 100
+    c = 0.9 ** (np.arange(n)**2)
+    y = random.randn(n)
+
+    solution1 = solve_toeplitz(c, b=y)
+    solution2 = solve(toeplitz(c), y)
+
+    assert_allclose(solution1, solution2)
+
+
+@pytest.mark.parametrize('dt_c', [int, float, np.float32, complex, np.complex64])
+@pytest.mark.parametrize('dt_b', [int, float, np.float32, complex, np.complex64])
+def test_empty(dt_c, dt_b):
+    c = np.array([], dtype=dt_c)
+    b = np.array([], dtype=dt_b)
+    x = solve_toeplitz(c, b)
+    assert x.shape == (0,)
+    assert x.dtype == solve_toeplitz(np.array([2, 1], dtype=dt_c),
+                                      np.ones(2, dtype=dt_b)).dtype
+
+    b = np.empty((0, 0), dtype=dt_b)
+    x1 = solve_toeplitz(c, b)
+    assert x1.shape == (0, 0)
+    assert x1.dtype == x.dtype
+
+
+@pytest.mark.parametrize('fun', [solve_toeplitz, matmul_toeplitz])
+def test_nd_FutureWarning(fun):
+    # Test future warnings with n-D `c`/`r`
+    rng = np.random.default_rng(283592436523456)
+    c = rng.random((2, 3, 4))
+    r = rng.random((2, 3, 4))
+    b_or_x = rng.random(24)
+    message = "Beginning in SciPy 1.17, multidimensional input will be..."
+    with pytest.warns(FutureWarning, match=message):
+         fun(c, b_or_x)
+    with pytest.warns(FutureWarning, match=message):
+         fun((c, r), b_or_x)