def _matmul(self, x, adjoint=False, adjoint_arg=False):
return _linalg.matmul(self._matrix,
x,
adjoint_a=adjoint,
adjoint_b=adjoint_arg)
def _solve(self, rhs, adjoint=False, adjoint_arg=False):
return self._dense_solve(rhs, adjoint=adjoint, adjoint_arg=adjoint_arg)
def _to_dense(self):
return self._matrix
@property
def _composite_tensor_fields(self):
return ("matrix", )
import numpy as np
from tensorflow_probability.python.internal.backend.numpy import linalg_impl as _linalg
from tensorflow_probability.python.internal.backend.numpy import ops as _ops
from tensorflow_probability.python.internal.backend.numpy.gen import tensor_shape
from tensorflow_probability.python.internal.backend.numpy import private
distribution_util = private.LazyLoader(
"distribution_util", globals(),
"tensorflow_probability.substrates.numpy.internal.distribution_util")
tensorshape_util = private.LazyLoader(
"tensorshape_util", globals(),
"tensorflow_probability.substrates.numpy.internal.tensorshape_util")
return _linalg.set_diag(mat, new_diag)
def _eigvals(self):
return self._ones_diag() * self.multiplier[..., _ops.newaxis]
def _cond(self):
# Condition number for a scalar time identity matrix is one, except when the
# scalar is zero.
return array_ops.where_v2(math_ops.equal(self._multiplier, 0.),
_ops.cast(np.nan, dtype=self.dtype),
_ops.cast(1., dtype=self.dtype))
@property
def multiplier(self):
"""The [batch] scalar `Tensor`, `c` in `cI`."""
return self._multiplier
import numpy as np
from tensorflow_probability.python.internal.backend.numpy import linalg_impl as _linalg
from tensorflow_probability.python.internal.backend.numpy import ops as _ops
from tensorflow_probability.python.internal.backend.numpy.gen import tensor_shape
from tensorflow_probability.python.internal.backend.numpy import private
distribution_util = private.LazyLoader(
"distribution_util", globals(),
"tensorflow_probability.python.internal._numpy.distribution_util")
tensorshape_util = private.LazyLoader(
"tensorshape_util", globals(),
"tensorflow_probability.python.internal._numpy.tensorshape_util")