708 lines
22 KiB
Python
708 lines
22 KiB
Python
from sympy.core.evalf import N
|
|
from sympy.core.numbers import (Float, I, Rational)
|
|
from sympy.core.symbol import (Symbol, symbols)
|
|
from sympy.functions.elementary.complexes import Abs
|
|
from sympy.functions.elementary.exponential import exp
|
|
from sympy.functions.elementary.miscellaneous import sqrt
|
|
from sympy.functions.elementary.trigonometric import (cos, sin)
|
|
from sympy.matrices import eye, Matrix
|
|
from sympy.core.singleton import S
|
|
from sympy.testing.pytest import raises, XFAIL
|
|
from sympy.matrices.matrices import NonSquareMatrixError, MatrixError
|
|
from sympy.matrices.expressions.fourier import DFT
|
|
from sympy.simplify.simplify import simplify
|
|
from sympy.matrices.immutable import ImmutableMatrix
|
|
from sympy.testing.pytest import slow
|
|
from sympy.testing.matrices import allclose
|
|
|
|
|
|
def test_eigen():
|
|
R = Rational
|
|
M = Matrix.eye(3)
|
|
assert M.eigenvals(multiple=False) == {S.One: 3}
|
|
assert M.eigenvals(multiple=True) == [1, 1, 1]
|
|
|
|
assert M.eigenvects() == (
|
|
[(1, 3, [Matrix([1, 0, 0]),
|
|
Matrix([0, 1, 0]),
|
|
Matrix([0, 0, 1])])])
|
|
|
|
assert M.left_eigenvects() == (
|
|
[(1, 3, [Matrix([[1, 0, 0]]),
|
|
Matrix([[0, 1, 0]]),
|
|
Matrix([[0, 0, 1]])])])
|
|
|
|
M = Matrix([[0, 1, 1],
|
|
[1, 0, 0],
|
|
[1, 1, 1]])
|
|
|
|
assert M.eigenvals() == {2*S.One: 1, -S.One: 1, S.Zero: 1}
|
|
|
|
assert M.eigenvects() == (
|
|
[
|
|
(-1, 1, [Matrix([-1, 1, 0])]),
|
|
( 0, 1, [Matrix([0, -1, 1])]),
|
|
( 2, 1, [Matrix([R(2, 3), R(1, 3), 1])])
|
|
])
|
|
|
|
assert M.left_eigenvects() == (
|
|
[
|
|
(-1, 1, [Matrix([[-2, 1, 1]])]),
|
|
(0, 1, [Matrix([[-1, -1, 1]])]),
|
|
(2, 1, [Matrix([[1, 1, 1]])])
|
|
])
|
|
|
|
a = Symbol('a')
|
|
M = Matrix([[a, 0],
|
|
[0, 1]])
|
|
|
|
assert M.eigenvals() == {a: 1, S.One: 1}
|
|
|
|
M = Matrix([[1, -1],
|
|
[1, 3]])
|
|
assert M.eigenvects() == ([(2, 2, [Matrix(2, 1, [-1, 1])])])
|
|
assert M.left_eigenvects() == ([(2, 2, [Matrix([[1, 1]])])])
|
|
|
|
M = Matrix([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
|
|
a = R(15, 2)
|
|
b = 3*33**R(1, 2)
|
|
c = R(13, 2)
|
|
d = (R(33, 8) + 3*b/8)
|
|
e = (R(33, 8) - 3*b/8)
|
|
|
|
def NS(e, n):
|
|
return str(N(e, n))
|
|
r = [
|
|
(a - b/2, 1, [Matrix([(12 + 24/(c - b/2))/((c - b/2)*e) + 3/(c - b/2),
|
|
(6 + 12/(c - b/2))/e, 1])]),
|
|
( 0, 1, [Matrix([1, -2, 1])]),
|
|
(a + b/2, 1, [Matrix([(12 + 24/(c + b/2))/((c + b/2)*d) + 3/(c + b/2),
|
|
(6 + 12/(c + b/2))/d, 1])]),
|
|
]
|
|
r1 = [(NS(r[i][0], 2), NS(r[i][1], 2),
|
|
[NS(j, 2) for j in r[i][2][0]]) for i in range(len(r))]
|
|
r = M.eigenvects()
|
|
r2 = [(NS(r[i][0], 2), NS(r[i][1], 2),
|
|
[NS(j, 2) for j in r[i][2][0]]) for i in range(len(r))]
|
|
assert sorted(r1) == sorted(r2)
|
|
|
|
eps = Symbol('eps', real=True)
|
|
|
|
M = Matrix([[abs(eps), I*eps ],
|
|
[-I*eps, abs(eps) ]])
|
|
|
|
assert M.eigenvects() == (
|
|
[
|
|
( 0, 1, [Matrix([[-I*eps/abs(eps)], [1]])]),
|
|
( 2*abs(eps), 1, [ Matrix([[I*eps/abs(eps)], [1]]) ] ),
|
|
])
|
|
|
|
assert M.left_eigenvects() == (
|
|
[
|
|
(0, 1, [Matrix([[I*eps/Abs(eps), 1]])]),
|
|
(2*Abs(eps), 1, [Matrix([[-I*eps/Abs(eps), 1]])])
|
|
])
|
|
|
|
M = Matrix(3, 3, [1, 2, 0, 0, 3, 0, 2, -4, 2])
|
|
M._eigenvects = M.eigenvects(simplify=False)
|
|
assert max(i.q for i in M._eigenvects[0][2][0]) > 1
|
|
M._eigenvects = M.eigenvects(simplify=True)
|
|
assert max(i.q for i in M._eigenvects[0][2][0]) == 1
|
|
|
|
M = Matrix([[Rational(1, 4), 1], [1, 1]])
|
|
assert M.eigenvects() == [
|
|
(Rational(5, 8) - sqrt(73)/8, 1, [Matrix([[-sqrt(73)/8 - Rational(3, 8)], [1]])]),
|
|
(Rational(5, 8) + sqrt(73)/8, 1, [Matrix([[Rational(-3, 8) + sqrt(73)/8], [1]])])]
|
|
|
|
# issue 10719
|
|
assert Matrix([]).eigenvals() == {}
|
|
assert Matrix([]).eigenvals(multiple=True) == []
|
|
assert Matrix([]).eigenvects() == []
|
|
|
|
# issue 15119
|
|
raises(NonSquareMatrixError,
|
|
lambda: Matrix([[1, 2], [0, 4], [0, 0]]).eigenvals())
|
|
raises(NonSquareMatrixError,
|
|
lambda: Matrix([[1, 0], [3, 4], [5, 6]]).eigenvals())
|
|
raises(NonSquareMatrixError,
|
|
lambda: Matrix([[1, 2, 3], [0, 5, 6]]).eigenvals())
|
|
raises(NonSquareMatrixError,
|
|
lambda: Matrix([[1, 0, 0], [4, 5, 0]]).eigenvals())
|
|
raises(NonSquareMatrixError,
|
|
lambda: Matrix([[1, 2, 3], [0, 5, 6]]).eigenvals(
|
|
error_when_incomplete = False))
|
|
raises(NonSquareMatrixError,
|
|
lambda: Matrix([[1, 0, 0], [4, 5, 0]]).eigenvals(
|
|
error_when_incomplete = False))
|
|
|
|
m = Matrix([[1, 2], [3, 4]])
|
|
assert isinstance(m.eigenvals(simplify=True, multiple=False), dict)
|
|
assert isinstance(m.eigenvals(simplify=True, multiple=True), list)
|
|
assert isinstance(m.eigenvals(simplify=lambda x: x, multiple=False), dict)
|
|
assert isinstance(m.eigenvals(simplify=lambda x: x, multiple=True), list)
|
|
|
|
|
|
@slow
|
|
def test_eigen_slow():
|
|
# issue 15125
|
|
from sympy.core.function import count_ops
|
|
q = Symbol("q", positive = True)
|
|
m = Matrix([[-2, exp(-q), 1], [exp(q), -2, 1], [1, 1, -2]])
|
|
assert count_ops(m.eigenvals(simplify=False)) > \
|
|
count_ops(m.eigenvals(simplify=True))
|
|
assert count_ops(m.eigenvals(simplify=lambda x: x)) > \
|
|
count_ops(m.eigenvals(simplify=True))
|
|
|
|
|
|
def test_float_eigenvals():
|
|
m = Matrix([[1, .6, .6], [.6, .9, .9], [.9, .6, .6]])
|
|
evals = [
|
|
Rational(5, 4) - sqrt(385)/20,
|
|
sqrt(385)/20 + Rational(5, 4),
|
|
S.Zero]
|
|
|
|
n_evals = m.eigenvals(rational=True, multiple=True)
|
|
n_evals = sorted(n_evals)
|
|
s_evals = [x.evalf() for x in evals]
|
|
s_evals = sorted(s_evals)
|
|
|
|
for x, y in zip(n_evals, s_evals):
|
|
assert abs(x-y) < 10**-9
|
|
|
|
|
|
@XFAIL
|
|
def test_eigen_vects():
|
|
m = Matrix(2, 2, [1, 0, 0, I])
|
|
raises(NotImplementedError, lambda: m.is_diagonalizable(True))
|
|
# !!! bug because of eigenvects() or roots(x**2 + (-1 - I)*x + I, x)
|
|
# see issue 5292
|
|
assert not m.is_diagonalizable(True)
|
|
raises(MatrixError, lambda: m.diagonalize(True))
|
|
(P, D) = m.diagonalize(True)
|
|
|
|
def test_issue_8240():
|
|
# Eigenvalues of large triangular matrices
|
|
x, y = symbols('x y')
|
|
n = 200
|
|
|
|
diagonal_variables = [Symbol('x%s' % i) for i in range(n)]
|
|
M = [[0 for i in range(n)] for j in range(n)]
|
|
for i in range(n):
|
|
M[i][i] = diagonal_variables[i]
|
|
M = Matrix(M)
|
|
|
|
eigenvals = M.eigenvals()
|
|
assert len(eigenvals) == n
|
|
for i in range(n):
|
|
assert eigenvals[diagonal_variables[i]] == 1
|
|
|
|
eigenvals = M.eigenvals(multiple=True)
|
|
assert set(eigenvals) == set(diagonal_variables)
|
|
|
|
# with multiplicity
|
|
M = Matrix([[x, 0, 0], [1, y, 0], [2, 3, x]])
|
|
eigenvals = M.eigenvals()
|
|
assert eigenvals == {x: 2, y: 1}
|
|
|
|
eigenvals = M.eigenvals(multiple=True)
|
|
assert len(eigenvals) == 3
|
|
assert eigenvals.count(x) == 2
|
|
assert eigenvals.count(y) == 1
|
|
|
|
|
|
def test_eigenvals():
|
|
M = Matrix([[0, 1, 1],
|
|
[1, 0, 0],
|
|
[1, 1, 1]])
|
|
assert M.eigenvals() == {2*S.One: 1, -S.One: 1, S.Zero: 1}
|
|
|
|
m = Matrix([
|
|
[3, 0, 0, 0, -3],
|
|
[0, -3, -3, 0, 3],
|
|
[0, 3, 0, 3, 0],
|
|
[0, 0, 3, 0, 3],
|
|
[3, 0, 0, 3, 0]])
|
|
|
|
# XXX Used dry-run test because arbitrary symbol that appears in
|
|
# CRootOf may not be unique.
|
|
assert m.eigenvals()
|
|
|
|
|
|
def test_eigenvects():
|
|
M = Matrix([[0, 1, 1],
|
|
[1, 0, 0],
|
|
[1, 1, 1]])
|
|
vecs = M.eigenvects()
|
|
for val, mult, vec_list in vecs:
|
|
assert len(vec_list) == 1
|
|
assert M*vec_list[0] == val*vec_list[0]
|
|
|
|
|
|
def test_left_eigenvects():
|
|
M = Matrix([[0, 1, 1],
|
|
[1, 0, 0],
|
|
[1, 1, 1]])
|
|
vecs = M.left_eigenvects()
|
|
for val, mult, vec_list in vecs:
|
|
assert len(vec_list) == 1
|
|
assert vec_list[0]*M == val*vec_list[0]
|
|
|
|
|
|
@slow
|
|
def test_bidiagonalize():
|
|
M = Matrix([[1, 0, 0],
|
|
[0, 1, 0],
|
|
[0, 0, 1]])
|
|
assert M.bidiagonalize() == M
|
|
assert M.bidiagonalize(upper=False) == M
|
|
assert M.bidiagonalize() == M
|
|
assert M.bidiagonal_decomposition() == (M, M, M)
|
|
assert M.bidiagonal_decomposition(upper=False) == (M, M, M)
|
|
assert M.bidiagonalize() == M
|
|
|
|
import random
|
|
#Real Tests
|
|
for real_test in range(2):
|
|
test_values = []
|
|
row = 2
|
|
col = 2
|
|
for _ in range(row * col):
|
|
value = random.randint(-1000000000, 1000000000)
|
|
test_values = test_values + [value]
|
|
# L -> Lower Bidiagonalization
|
|
# M -> Mutable Matrix
|
|
# N -> Immutable Matrix
|
|
# 0 -> Bidiagonalized form
|
|
# 1,2,3 -> Bidiagonal_decomposition matrices
|
|
# 4 -> Product of 1 2 3
|
|
M = Matrix(row, col, test_values)
|
|
N = ImmutableMatrix(M)
|
|
|
|
N1, N2, N3 = N.bidiagonal_decomposition()
|
|
M1, M2, M3 = M.bidiagonal_decomposition()
|
|
M0 = M.bidiagonalize()
|
|
N0 = N.bidiagonalize()
|
|
|
|
N4 = N1 * N2 * N3
|
|
M4 = M1 * M2 * M3
|
|
|
|
N2.simplify()
|
|
N4.simplify()
|
|
N0.simplify()
|
|
|
|
M0.simplify()
|
|
M2.simplify()
|
|
M4.simplify()
|
|
|
|
LM0 = M.bidiagonalize(upper=False)
|
|
LM1, LM2, LM3 = M.bidiagonal_decomposition(upper=False)
|
|
LN0 = N.bidiagonalize(upper=False)
|
|
LN1, LN2, LN3 = N.bidiagonal_decomposition(upper=False)
|
|
|
|
LN4 = LN1 * LN2 * LN3
|
|
LM4 = LM1 * LM2 * LM3
|
|
|
|
LN2.simplify()
|
|
LN4.simplify()
|
|
LN0.simplify()
|
|
|
|
LM0.simplify()
|
|
LM2.simplify()
|
|
LM4.simplify()
|
|
|
|
assert M == M4
|
|
assert M2 == M0
|
|
assert N == N4
|
|
assert N2 == N0
|
|
assert M == LM4
|
|
assert LM2 == LM0
|
|
assert N == LN4
|
|
assert LN2 == LN0
|
|
|
|
#Complex Tests
|
|
for complex_test in range(2):
|
|
test_values = []
|
|
size = 2
|
|
for _ in range(size * size):
|
|
real = random.randint(-1000000000, 1000000000)
|
|
comp = random.randint(-1000000000, 1000000000)
|
|
value = real + comp * I
|
|
test_values = test_values + [value]
|
|
M = Matrix(size, size, test_values)
|
|
N = ImmutableMatrix(M)
|
|
# L -> Lower Bidiagonalization
|
|
# M -> Mutable Matrix
|
|
# N -> Immutable Matrix
|
|
# 0 -> Bidiagonalized form
|
|
# 1,2,3 -> Bidiagonal_decomposition matrices
|
|
# 4 -> Product of 1 2 3
|
|
N1, N2, N3 = N.bidiagonal_decomposition()
|
|
M1, M2, M3 = M.bidiagonal_decomposition()
|
|
M0 = M.bidiagonalize()
|
|
N0 = N.bidiagonalize()
|
|
|
|
N4 = N1 * N2 * N3
|
|
M4 = M1 * M2 * M3
|
|
|
|
N2.simplify()
|
|
N4.simplify()
|
|
N0.simplify()
|
|
|
|
M0.simplify()
|
|
M2.simplify()
|
|
M4.simplify()
|
|
|
|
LM0 = M.bidiagonalize(upper=False)
|
|
LM1, LM2, LM3 = M.bidiagonal_decomposition(upper=False)
|
|
LN0 = N.bidiagonalize(upper=False)
|
|
LN1, LN2, LN3 = N.bidiagonal_decomposition(upper=False)
|
|
|
|
LN4 = LN1 * LN2 * LN3
|
|
LM4 = LM1 * LM2 * LM3
|
|
|
|
LN2.simplify()
|
|
LN4.simplify()
|
|
LN0.simplify()
|
|
|
|
LM0.simplify()
|
|
LM2.simplify()
|
|
LM4.simplify()
|
|
|
|
assert M == M4
|
|
assert M2 == M0
|
|
assert N == N4
|
|
assert N2 == N0
|
|
assert M == LM4
|
|
assert LM2 == LM0
|
|
assert N == LN4
|
|
assert LN2 == LN0
|
|
|
|
M = Matrix(18, 8, range(1, 145))
|
|
M = M.applyfunc(lambda i: Float(i))
|
|
assert M.bidiagonal_decomposition()[1] == M.bidiagonalize()
|
|
assert M.bidiagonal_decomposition(upper=False)[1] == M.bidiagonalize(upper=False)
|
|
a, b, c = M.bidiagonal_decomposition()
|
|
diff = a * b * c - M
|
|
assert abs(max(diff)) < 10**-12
|
|
|
|
|
|
def test_diagonalize():
|
|
m = Matrix(2, 2, [0, -1, 1, 0])
|
|
raises(MatrixError, lambda: m.diagonalize(reals_only=True))
|
|
P, D = m.diagonalize()
|
|
assert D.is_diagonal()
|
|
assert D == Matrix([
|
|
[-I, 0],
|
|
[ 0, I]])
|
|
|
|
# make sure we use floats out if floats are passed in
|
|
m = Matrix(2, 2, [0, .5, .5, 0])
|
|
P, D = m.diagonalize()
|
|
assert all(isinstance(e, Float) for e in D.values())
|
|
assert all(isinstance(e, Float) for e in P.values())
|
|
|
|
_, D2 = m.diagonalize(reals_only=True)
|
|
assert D == D2
|
|
|
|
m = Matrix(
|
|
[[0, 1, 0, 0], [1, 0, 0, 0.002], [0.002, 0, 0, 1], [0, 0, 1, 0]])
|
|
P, D = m.diagonalize()
|
|
assert allclose(P*D, m*P)
|
|
|
|
|
|
def test_is_diagonalizable():
|
|
a, b, c = symbols('a b c')
|
|
m = Matrix(2, 2, [a, c, c, b])
|
|
assert m.is_symmetric()
|
|
assert m.is_diagonalizable()
|
|
assert not Matrix(2, 2, [1, 1, 0, 1]).is_diagonalizable()
|
|
|
|
m = Matrix(2, 2, [0, -1, 1, 0])
|
|
assert m.is_diagonalizable()
|
|
assert not m.is_diagonalizable(reals_only=True)
|
|
|
|
|
|
def test_jordan_form():
|
|
m = Matrix(3, 2, [-3, 1, -3, 20, 3, 10])
|
|
raises(NonSquareMatrixError, lambda: m.jordan_form())
|
|
|
|
# the next two tests test the cases where the old
|
|
# algorithm failed due to the fact that the block structure can
|
|
# *NOT* be determined from algebraic and geometric multiplicity alone
|
|
# This can be seen most easily when one lets compute the J.c.f. of a matrix that
|
|
# is in J.c.f already.
|
|
m = Matrix(4, 4, [2, 1, 0, 0,
|
|
0, 2, 1, 0,
|
|
0, 0, 2, 0,
|
|
0, 0, 0, 2
|
|
])
|
|
P, J = m.jordan_form()
|
|
assert m == J
|
|
|
|
m = Matrix(4, 4, [2, 1, 0, 0,
|
|
0, 2, 0, 0,
|
|
0, 0, 2, 1,
|
|
0, 0, 0, 2
|
|
])
|
|
P, J = m.jordan_form()
|
|
assert m == J
|
|
|
|
A = Matrix([[ 2, 4, 1, 0],
|
|
[-4, 2, 0, 1],
|
|
[ 0, 0, 2, 4],
|
|
[ 0, 0, -4, 2]])
|
|
P, J = A.jordan_form()
|
|
assert simplify(P*J*P.inv()) == A
|
|
|
|
assert Matrix(1, 1, [1]).jordan_form() == (Matrix([1]), Matrix([1]))
|
|
assert Matrix(1, 1, [1]).jordan_form(calc_transform=False) == Matrix([1])
|
|
|
|
# If we have eigenvalues in CRootOf form, raise errors
|
|
m = Matrix([[3, 0, 0, 0, -3], [0, -3, -3, 0, 3], [0, 3, 0, 3, 0], [0, 0, 3, 0, 3], [3, 0, 0, 3, 0]])
|
|
raises(MatrixError, lambda: m.jordan_form())
|
|
|
|
# make sure that if the input has floats, the output does too
|
|
m = Matrix([
|
|
[ 0.6875, 0.125 + 0.1875*sqrt(3)],
|
|
[0.125 + 0.1875*sqrt(3), 0.3125]])
|
|
P, J = m.jordan_form()
|
|
assert all(isinstance(x, Float) or x == 0 for x in P)
|
|
assert all(isinstance(x, Float) or x == 0 for x in J)
|
|
|
|
|
|
def test_singular_values():
|
|
x = Symbol('x', real=True)
|
|
|
|
A = Matrix([[0, 1*I], [2, 0]])
|
|
# if singular values can be sorted, they should be in decreasing order
|
|
assert A.singular_values() == [2, 1]
|
|
|
|
A = eye(3)
|
|
A[1, 1] = x
|
|
A[2, 2] = 5
|
|
vals = A.singular_values()
|
|
# since Abs(x) cannot be sorted, test set equality
|
|
assert set(vals) == {5, 1, Abs(x)}
|
|
|
|
A = Matrix([[sin(x), cos(x)], [-cos(x), sin(x)]])
|
|
vals = [sv.trigsimp() for sv in A.singular_values()]
|
|
assert vals == [S.One, S.One]
|
|
|
|
A = Matrix([
|
|
[2, 4],
|
|
[1, 3],
|
|
[0, 0],
|
|
[0, 0]
|
|
])
|
|
assert A.singular_values() == \
|
|
[sqrt(sqrt(221) + 15), sqrt(15 - sqrt(221))]
|
|
assert A.T.singular_values() == \
|
|
[sqrt(sqrt(221) + 15), sqrt(15 - sqrt(221)), 0, 0]
|
|
|
|
def test___eq__():
|
|
assert (Matrix(
|
|
[[0, 1, 1],
|
|
[1, 0, 0],
|
|
[1, 1, 1]]) == {}) is False
|
|
|
|
|
|
def test_definite():
|
|
# Examples from Gilbert Strang, "Introduction to Linear Algebra"
|
|
# Positive definite matrices
|
|
m = Matrix([[2, -1, 0], [-1, 2, -1], [0, -1, 2]])
|
|
assert m.is_positive_definite == True
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == False
|
|
|
|
m = Matrix([[5, 4], [4, 5]])
|
|
assert m.is_positive_definite == True
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == False
|
|
|
|
# Positive semidefinite matrices
|
|
m = Matrix([[2, -1, -1], [-1, 2, -1], [-1, -1, 2]])
|
|
assert m.is_positive_definite == False
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == False
|
|
|
|
m = Matrix([[1, 2], [2, 4]])
|
|
assert m.is_positive_definite == False
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == False
|
|
|
|
# Examples from Mathematica documentation
|
|
# Non-hermitian positive definite matrices
|
|
m = Matrix([[2, 3], [4, 8]])
|
|
assert m.is_positive_definite == True
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == False
|
|
|
|
# Hermetian matrices
|
|
m = Matrix([[1, 2*I], [-I, 4]])
|
|
assert m.is_positive_definite == True
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == False
|
|
|
|
# Symbolic matrices examples
|
|
a = Symbol('a', positive=True)
|
|
b = Symbol('b', negative=True)
|
|
m = Matrix([[a, 0, 0], [0, a, 0], [0, 0, a]])
|
|
assert m.is_positive_definite == True
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == False
|
|
|
|
m = Matrix([[b, 0, 0], [0, b, 0], [0, 0, b]])
|
|
assert m.is_positive_definite == False
|
|
assert m.is_positive_semidefinite == False
|
|
assert m.is_negative_definite == True
|
|
assert m.is_negative_semidefinite == True
|
|
assert m.is_indefinite == False
|
|
|
|
m = Matrix([[a, 0], [0, b]])
|
|
assert m.is_positive_definite == False
|
|
assert m.is_positive_semidefinite == False
|
|
assert m.is_negative_definite == False
|
|
assert m.is_negative_semidefinite == False
|
|
assert m.is_indefinite == True
|
|
|
|
m = Matrix([
|
|
[0.0228202735623867, 0.00518748979085398,
|
|
-0.0743036351048907, -0.00709135324903921],
|
|
[0.00518748979085398, 0.0349045359786350,
|
|
0.0830317991056637, 0.00233147902806909],
|
|
[-0.0743036351048907, 0.0830317991056637,
|
|
1.15859676366277, 0.340359081555988],
|
|
[-0.00709135324903921, 0.00233147902806909,
|
|
0.340359081555988, 0.928147644848199]
|
|
])
|
|
assert m.is_positive_definite == True
|
|
assert m.is_positive_semidefinite == True
|
|
assert m.is_indefinite == False
|
|
|
|
# test for issue 19547: https://github.com/sympy/sympy/issues/19547
|
|
m = Matrix([
|
|
[0, 0, 0],
|
|
[0, 1, 2],
|
|
[0, 2, 1]
|
|
])
|
|
assert not m.is_positive_definite
|
|
assert not m.is_positive_semidefinite
|
|
|
|
|
|
def test_positive_semidefinite_cholesky():
|
|
from sympy.matrices.eigen import _is_positive_semidefinite_cholesky
|
|
|
|
m = Matrix([[0, 0, 0], [0, 0, 0], [0, 0, 0]])
|
|
assert _is_positive_semidefinite_cholesky(m) == True
|
|
m = Matrix([[0, 0, 0], [0, 5, -10*I], [0, 10*I, 5]])
|
|
assert _is_positive_semidefinite_cholesky(m) == False
|
|
m = Matrix([[1, 0, 0], [0, 0, 0], [0, 0, -1]])
|
|
assert _is_positive_semidefinite_cholesky(m) == False
|
|
m = Matrix([[0, 1], [1, 0]])
|
|
assert _is_positive_semidefinite_cholesky(m) == False
|
|
|
|
# https://www.value-at-risk.net/cholesky-factorization/
|
|
m = Matrix([[4, -2, -6], [-2, 10, 9], [-6, 9, 14]])
|
|
assert _is_positive_semidefinite_cholesky(m) == True
|
|
m = Matrix([[9, -3, 3], [-3, 2, 1], [3, 1, 6]])
|
|
assert _is_positive_semidefinite_cholesky(m) == True
|
|
m = Matrix([[4, -2, 2], [-2, 1, -1], [2, -1, 5]])
|
|
assert _is_positive_semidefinite_cholesky(m) == True
|
|
m = Matrix([[1, 2, -1], [2, 5, 1], [-1, 1, 9]])
|
|
assert _is_positive_semidefinite_cholesky(m) == False
|
|
|
|
|
|
def test_issue_20582():
|
|
A = Matrix([
|
|
[5, -5, -3, 2, -7],
|
|
[-2, -5, 0, 2, 1],
|
|
[-2, -7, -5, -2, -6],
|
|
[7, 10, 3, 9, -2],
|
|
[4, -10, 3, -8, -4]
|
|
])
|
|
# XXX Used dry-run test because arbitrary symbol that appears in
|
|
# CRootOf may not be unique.
|
|
assert A.eigenvects()
|
|
|
|
def test_issue_19210():
|
|
t = Symbol('t')
|
|
H = Matrix([[3, 0, 0, 0], [0, 1 , 2, 0], [0, 2, 2, 0], [0, 0, 0, 4]])
|
|
A = (-I * H * t).jordan_form()
|
|
assert A == (Matrix([
|
|
[0, 1, 0, 0],
|
|
[0, 0, -4/(-1 + sqrt(17)), 4/(1 + sqrt(17))],
|
|
[0, 0, 1, 1],
|
|
[1, 0, 0, 0]]), Matrix([
|
|
[-4*I*t, 0, 0, 0],
|
|
[ 0, -3*I*t, 0, 0],
|
|
[ 0, 0, t*(-3*I/2 + sqrt(17)*I/2), 0],
|
|
[ 0, 0, 0, t*(-sqrt(17)*I/2 - 3*I/2)]]))
|
|
|
|
|
|
def test_issue_20275():
|
|
# XXX We use complex expansions because complex exponentials are not
|
|
# recognized by polys.domains
|
|
A = DFT(3).as_explicit().expand(complex=True)
|
|
eigenvects = A.eigenvects()
|
|
assert eigenvects[0] == (
|
|
-1, 1,
|
|
[Matrix([[1 - sqrt(3)], [1], [1]])]
|
|
)
|
|
assert eigenvects[1] == (
|
|
1, 1,
|
|
[Matrix([[1 + sqrt(3)], [1], [1]])]
|
|
)
|
|
assert eigenvects[2] == (
|
|
-I, 1,
|
|
[Matrix([[0], [-1], [1]])]
|
|
)
|
|
|
|
A = DFT(4).as_explicit().expand(complex=True)
|
|
eigenvects = A.eigenvects()
|
|
assert eigenvects[0] == (
|
|
-1, 1,
|
|
[Matrix([[-1], [1], [1], [1]])]
|
|
)
|
|
assert eigenvects[1] == (
|
|
1, 2,
|
|
[Matrix([[1], [0], [1], [0]]), Matrix([[2], [1], [0], [1]])]
|
|
)
|
|
assert eigenvects[2] == (
|
|
-I, 1,
|
|
[Matrix([[0], [-1], [0], [1]])]
|
|
)
|
|
|
|
# XXX We skip test for some parts of eigenvectors which are very
|
|
# complicated and fragile under expression tree changes
|
|
A = DFT(5).as_explicit().expand(complex=True)
|
|
eigenvects = A.eigenvects()
|
|
assert eigenvects[0] == (
|
|
-1, 1,
|
|
[Matrix([[1 - sqrt(5)], [1], [1], [1], [1]])]
|
|
)
|
|
assert eigenvects[1] == (
|
|
1, 2,
|
|
[Matrix([[S(1)/2 + sqrt(5)/2], [0], [1], [1], [0]]),
|
|
Matrix([[S(1)/2 + sqrt(5)/2], [1], [0], [0], [1]])]
|
|
)
|
|
|
|
|
|
def test_issue_20752():
|
|
b = symbols('b', nonzero=True)
|
|
m = Matrix([[0, 0, 0], [0, b, 0], [0, 0, b]])
|
|
assert m.is_positive_semidefinite is None
|