def test_doit_args():
A = ImmutableMatrix([[1, 2], [3, 4]])
B = ImmutableMatrix([[2, 3], [4, 5]])
assert MatAdd(A, MatPow(B, 2)).doit() == A + B**2
assert MatAdd(A, MatMul(A, B)).doit() == A + A*B
assert (MatAdd(A, X, MatMul(A, B), Y, MatAdd(2*A, B)).doit() ==
MatAdd(3*A + A*B + B, X, Y))
def _eval_derivative_matrix_lines(self, x):
from sympy.core.expr import ExprBuilder
from sympy.codegen.array_utils import (
CodegenArrayContraction,
CodegenArrayTensorProduct,
)
from .matmul import MatMul
from .inverse import Inverse
exp = self.exp
if self.base.shape == (1, 1) and not exp.has(x):
lr = self.base._eval_derivative_matrix_lines(x)
for i in lr:
subexpr = ExprBuilder(
CodegenArrayContraction,
[
ExprBuilder(
CodegenArrayTensorProduct,
[
Identity(1),
i._lines[0],
exp * self.base**(exp - 1),
i._lines[1],
Identity(1),
],
),
(0, 3, 4),
(5, 7, 8),
],
validator=CodegenArrayContraction._validate,
)
i._first_pointer_parent = subexpr.args[0].args
i._first_pointer_index = 0
i._second_pointer_parent = subexpr.args[0].args
i._second_pointer_index = 4
i._lines = [subexpr]
return lr
if (exp > 0) == True:
newexpr = MatMul.fromiter([self.base for i in range(exp)])
elif (exp == -1) == True:
return Inverse(self.base)._eval_derivative_matrix_lines(x)
elif (exp < 0) == True:
newexpr = MatMul.fromiter(
[Inverse(self.base) for i in range(-exp)])
elif (exp == 0) == True:
return self.doit()._eval_derivative_matrix_lines(x)
else:
raise NotImplementedError("cannot evaluate %s derived by %s" %
(self, x))
return newexpr._eval_derivative_matrix_lines(x)
def _eval_derivative_matrix_lines(self, x):
from .matmul import MatMul
exp = self.exp
if (exp > 0) == True:
newexpr = MatMul.fromiter([self.base for i in range(exp)])
elif (exp == -1) == True:
return Inverse(self.base)._eval_derivative_matrix_lines(x)
elif (exp < 0) == True:
newexpr = MatMul.fromiter([Inverse(self.base) for i in range(-exp)])
elif (exp == 0) == True:
return self.doit()._eval_derivative_matrix_lines(x)
else:
raise NotImplementedError("cannot evaluate %s derived by %s" % (self, x))
return newexpr._eval_derivative_matrix_lines(x)
def _entry(self, i, j, **kwargs):
from sympy.matrices.expressions import MatMul
A = self.doit()
if isinstance(A, MatPow):
# We still have a MatPow, make an explicit MatMul out of it.
if A.exp.is_Integer and A.exp.is_positive:
A = MatMul(*[A.base for k in range(A.exp)])
elif not self._is_shape_symbolic():
return A._get_explicit_matrix()[i, j]
else:
# Leave the expression unevaluated:
from sympy.matrices.expressions.matexpr import MatrixElement
return MatrixElement(self, i, j)
return A[i, j]
def _eval_derivative_matrix_lines(self, x):
from .matmul import MatMul
from .inverse import Inverse
exp = self.exp
if (exp > 0) == True:
newexpr = MatMul.fromiter([self.base for i in range(exp)])
elif (exp == -1) == True:
return Inverse(self.base)._eval_derivative_matrix_lines(x)
elif (exp < 0) == True:
newexpr = MatMul.fromiter([Inverse(self.base) for i in range(-exp)])
elif (exp == 0) == True:
return self.doit()._eval_derivative_matrix_lines(x)
else:
raise NotImplementedError("cannot evaluate %s derived by %s" % (self, x))
return newexpr._eval_derivative_matrix_lines(x)
def test_Trace_doit_deep_False():
X = Matrix([[1, 2], [3, 4]])
q = MatPow(X, 2)
assert Trace(q).doit(deep=False).arg == q
q = MatAdd(X, 2 * X)
assert Trace(q).doit(deep=False).arg == q
q = MatMul(X, 2 * X)
assert Trace(q).doit(deep=False).arg == q
def _(expr, assumptions):
factor, mmul = expr.as_coeff_mmul()
if (all(ask(Q.positive_definite(arg), assumptions) for arg in mmul.args)
and factor > 0):
return True
if (len(mmul.args) >= 2 and mmul.args[0] == mmul.args[-1].T
and ask(Q.fullrank(mmul.args[0]), assumptions)):
return ask(Q.positive_definite(MatMul(*mmul.args[1:-1])), assumptions)
def MatMul(expr, assumptions):
factor, mmul = expr.as_coeff_mmul()
if all(ask(Q.symmetric(arg), assumptions) for arg in mmul.args):
return True
if len(mmul.args) >= 2 and mmul.args[0] == mmul.args[-1].T:
if len(mmul.args) == 2:
return True
return ask(Q.symmetric(MatMul(*mmul.args[1:-1])), assumptions)
def _eval_derivative_matrix_lines(self, x):
from sympy.core.expr import ExprBuilder
from sympy.codegen.array_utils import CodegenArrayContraction, CodegenArrayTensorProduct
from .matmul import MatMul
from .inverse import Inverse
exp = self.exp
if self.base.shape == (1, 1) and not exp.has(x):
lr = self.base._eval_derivative_matrix_lines(x)
for i in lr:
subexpr = ExprBuilder(
CodegenArrayContraction,
[
ExprBuilder(
CodegenArrayTensorProduct,
[
Identity(1),
i._lines[0],
exp*self.base**(exp-1),
i._lines[1],
Identity(1),
]
),
(0, 3, 4), (5, 7, 8)
],
validator=CodegenArrayContraction._validate
)
i._first_pointer_parent = subexpr.args[0].args
i._first_pointer_index = 0
i._second_pointer_parent = subexpr.args[0].args
i._second_pointer_index = 4
i._lines = [subexpr]
return lr
if (exp > 0) == True:
newexpr = MatMul.fromiter([self.base for i in range(exp)])
elif (exp == -1) == True:
return Inverse(self.base)._eval_derivative_matrix_lines(x)
elif (exp < 0) == True:
newexpr = MatMul.fromiter([Inverse(self.base) for i in range(-exp)])
elif (exp == 0) == True:
return self.doit()._eval_derivative_matrix_lines(x)
else:
raise NotImplementedError("cannot evaluate %s derived by %s" % (self, x))
return newexpr._eval_derivative_matrix_lines(x)
def _entry(self, i, j, **kwargs):
from sympy.matrices.expressions import MatMul
A = self.doit()
if isinstance(A, MatPow):
# We still have a MatPow, make an explicit MatMul out of it.
if not A.base.is_square:
raise ShapeError("Power of non-square matrix %s" % A.base)
elif A.exp.is_Integer and A.exp.is_positive:
A = MatMul(*[A.base for k in range(A.exp)])
#elif A.exp.is_Integer and self.exp.is_negative:
# Note: possible future improvement: in principle we can take
# positive powers of the inverse, but carefully avoid recursion,
# perhaps by adding `_entry` to Inverse (as it is our subclass).
# T = A.base.as_explicit().inverse()
# A = MatMul(*[T for k in range(-A.exp)])
else:
# Leave the expression unevaluated:
from sympy.matrices.expressions.matexpr import MatrixElement
return MatrixElement(self, i, j)
return A._entry(i, j)
def _entry(self, i, j, **kwargs):
from sympy.matrices.expressions import MatMul
A = self.doit()
if isinstance(A, MatPow):
# We still have a MatPow, make an explicit MatMul out of it.
if not A.base.is_square:
raise ShapeError("Power of non-square matrix %s" % A.base)
elif A.exp.is_Integer and A.exp.is_positive:
A = MatMul(*[A.base for k in range(A.exp)])
#elif A.exp.is_Integer and self.exp.is_negative:
# Note: possible future improvement: in principle we can take
# positive powers of the inverse, but carefully avoid recursion,
# perhaps by adding `_entry` to Inverse (as it is our subclass).
# T = A.base.as_explicit().inverse()
# A = MatMul(*[T for k in range(-A.exp)])
else:
# Leave the expression unevaluated:
from sympy.matrices.expressions.matexpr import MatrixElement
return MatrixElement(self, i, j)
return A._entry(i, j)
def MatMul(expr, assumptions):
factor, mmul = expr.as_coeff_mmul()
if all(ask(Q.symmetric(arg), assumptions) for arg in mmul.args):
return True
# TODO: implement sathandlers system for the matrices.
# Now it duplicates the general fact: Implies(Q.diagonal, Q.symmetric).
if ask(Q.diagonal(expr), assumptions):
return True
if len(mmul.args) >= 2 and mmul.args[0] == mmul.args[-1].T:
if len(mmul.args) == 2:
return True
return ask(Q.symmetric(MatMul(*mmul.args[1:-1])), assumptions)
def test_doit_nested_MatrixExpr():
X = ImmutableMatrix([[1, 2], [3, 4]])
Y = ImmutableMatrix([[2, 3], [4, 5]])
assert MatPow(MatMul(X, Y), 2).doit() == (X*Y)**2
assert MatPow(MatAdd(X, Y), 2).doit() == (X + Y)**2
def test_trace_constant_factor():
# Issue 9052: gave 2*Trace(MatMul(A)) instead of 2*Trace(A)
assert trace(2 * A) == 2 * Trace(A)
X = ImmutableMatrix([[1, 2], [3, 4]])
assert trace(MatMul(2, X)) == 10
def test_bc_matpow():
A = MatrixSymbol('A', n, n)
assert bc_matpow(A**2) == MatMul(A, A, evaluate=False)
assert bc_matpow(A**n) == A**n
def test_trace_constant_factor():
# Issue 9052: LHS gives Trace(MatMul(A))
assert trace(2 * A) == 2 * Trace(A)
X = ImmutableMatrix([[1, 2], [3, 4]])
assert trace(MatMul(2, X)) == 10
