Math

This submodule contains various mathematical functions. Most of them are imported directly from pytensor.tensor (see there for more details). Doing any kind of math with PyMC random variables, or defining custom likelihoods or priors requires you to use these PyTensor expressions rather than NumPy or Python code.

Functions exposed in pymc namespace

expand_packed_triangular(n, packed[, lower, ...])	Convert a packed triangular matrix into a two dimensional array.
logit(p)
invlogit	Logistic sigmoid function (1 / (1 + exp(-x)), also known as expit or inverse logit
probit(p)
invprobit(x)
logaddexp(*xs)	Logarithm of the sum of exponentiations of the inputs.
logsumexp(x[, axis, keepdims])	Compute the log of the sum of exponentials of input elements.

Functions exposed in pymc.math

abs	|`a`|
prod(input[, axis, dtype, keepdims, ...])	Computes the product along the given axis(es) of a tensor input.
dot(l, r)	Return a symbolic dot product.
eq	a == b
neq	a != b
ge	a >= b
gt	a > b
le	a <= b
lt	a < b
exp	e^`a`
log	base e logarithm of a
sgn(a)	sign of a
sqr	square of a
sqrt	square root of a
sum(input[, axis, dtype, keepdims, acc_dtype])	Computes the sum along the given axis(es) of a tensor input.
ceil	ceiling of a
floor	floor of a
sin	sine of a
sinh	hyperbolic sine of a
arcsin	arcsine of a
arcsinh	hyperbolic arc sine of a
cos	cosine of a
cosh	hyperbolic cosine of a
arccos	arccosine of a
arccosh	hyperbolic arc cosine of a
tan	tangent of a
tanh	hyperbolic tangent of a
arctan	arctangent of a
arctanh	hyperbolic arc tangent of a
cumprod(x[, axis])	Return the cumulative product of the elements along a given axis.
cumsum(x[, axis])	Return the cumulative sum of the elements along a given axis.
matmul(x1, x2[, dtype])	Compute the matrix product of two tensor variables.
and_	bitwise a & b
broadcast_to(x, shape)	Broadcast an array to a new shape.
clip	Clip x to be between min and max.
concatenate(tensor_list[, axis])	Alias for `join`(axis, *tensor_list).
flatten(x[, ndim])	Return a copy of the array collapsed into one dimension.
or_	bitwise a | b
stack(tensors[, axis])	Stack tensors in sequence on given axis (default is 0).
switch	if cond then ift else iff
where	if cond then ift else iff
flatten_list(tensors)
constant(x[, name, ndim, dtype])	Return a TensorConstant with value x.
max(x[, axis, keepdims])	Returns maximum elements obtained by iterating over given axis.
maximum	elemwise maximum.
mean(input[, axis, dtype, op, keepdims, ...])	Computes the mean value along the given axis(es) of a tensor input.
min(x[, axis, keepdims])	Returns minimum elements obtained by iterating over given axis.
minimum	elemwise minimum.
round(a[, mode])	round_mode(a) with mode in [half_away_from_zero, half_to_even].
erf	error function
erfc	complementary error function
erfcinv	inverse complementary error function
erfinv	inverse error function
log1pexp	Compute log(1 + exp(x)), also known as softplus or log1pexp
log1mexp(x, *[, negative_input])	Return log(1 - exp(-x)).
logaddexp(*xs)	Logarithm of the sum of exponentiations of the inputs.
logsumexp(x[, axis, keepdims])	Compute the log of the sum of exponentials of input elements.
logdiffexp(exp - exp)
logit(p)
invlogit	Logistic sigmoid function (1 / (1 + exp(-x)), also known as expit or inverse logit
probit(p)
invprobit(x)
sigmoid	Logistic sigmoid function (1 / (1 + exp(-x)), also known as expit or inverse logit
softmax(c[, axis])
log_softmax(c[, axis])
logbern(log_p)
full(shape, fill_value[, dtype])	Return a new array of given shape and type, filled with fill_value.
full_like(a, fill_value[, dtype])	Equivalent of numpy.full_like.
ones(shape[, dtype])	Create a TensorVariable filled with ones, closer to NumPy's syntax than alloc.
ones_like(model[, dtype, opt])	equivalent of numpy.ones_like Parameters ---------- model : tensor dtype : data-type, optional opt : If True, we will return a constant instead of a graph when possible. Useful for PyTensor optimization, not for user building a graph as this have the consequence that model isn't always in the graph.
zeros(shape[, dtype])	Create a TensorVariable filled with zeros, closer to NumPy's syntax than alloc.
zeros_like(model[, dtype, opt])	equivalent of numpy.zeros_like Parameters ---------- model : tensor dtype : data-type, optional opt : If True, we will return a constant instead of a graph when possible. Useful for PyTensor optimization, not for user building a graph as this have the consequence that model isn't always in the graph.
kronecker(*Ks)	Return the Kronecker product of arguments:
cartesian(*arrays)	Makes the Cartesian product of arrays.
kron_dot(krons, m, *[, op])	Apply op to krons and m in a way that reproduces op(kronecker(*krons), m)
kron_solve_lower(krons, m, *[, op])	Apply op to krons and m in a way that reproduces op(kronecker(*krons), m)
kron_solve_upper(krons, m, *[, op])	Apply op to krons and m in a way that reproduces op(kronecker(*krons), m)
kron_diag(*diags)	Returns diagonal of a kronecker product.
flat_outer(a, b)
expand_packed_triangular(n, packed[, lower, ...])	Convert a packed triangular matrix into a two dimensional array.
batched_diag(C)
block_diagonal(matrices[, sparse, format])	See pt.slinalg.block_diag or pytensor.sparse.basic.block_diag for reference
matrix_inverse	Generalizes a core Op to work with batched dimensions.
logdet	Compute the logarithm of the absolute determinant of a square matrix M, log(abs(det(M))) on the CPU.