def mom_call(self, k, dist):
"Moment generator call backend wrapper"
assert len(k) == len(dist)
graph = self.graph
self.dist, dist_ = dist, self.dist
graph.add_node(dist, key=k)
if hasattr(dist, "_mom"):
if dist.advance:
out = dist._mom(k, self)
else:
out = np.empty(k.shape[1:])
prm = self.D.build()
prm.update(self.K.build())
out[:] = dist._mom(k, **prm)
else:
out = mom(dist, k, **self.meta)
graph.add_node(dist, val=out)
self.dist = dist_
return np.array(out)
def mom_call(self, k, dist):
"Moment generator call backend wrapper"
assert len(k) == len(dist)
graph = self.graph
self.dist, dist_ = dist, self.dist
graph.add_node(dist, key=k)
if hasattr(dist, "_mom"):
if dist.advance:
out = dist._mom(k, self)
else:
out = np.empty(k.shape[1:])
prm = self.D.build()
prm.update(self.K.build())
out[:] = dist._mom(k, **prm)
else:
out = mom(dist, k, **self.meta)
graph.add_node(dist, val=out)
self.dist = dist_
return np.array(out)
def mom(self, K, **kws):
"""Raw statistical moments.
Creates non-centralized raw moments from the random variable. If
analytical options can not be utilized, Monte Carlo integration
will be used.
Parameters
----------
K : array_like
Index of the raw moments. k.shape must be compatible with
distribution shape.
Sampling scheme when performing Monte Carlo
rule : str
rule for estimating the moment if the analytical method
fails.
Key Monte Carlo schemes
--- ------------------------
"H" Halton sequence
"K" Korobov set
"L" Latin hypercube sampling
"M" Hammersley sequence
"R" (Pseudo-)Random sampling
"S" Sobol sequence
Key Quadrature schemes
--- ------------------------
"C" Clenshaw-Curtis
"Q" Gaussian quadrature
"E" Gauss-Legendre
composit : int, array_like optional
If provided, composit quadrature will be used.
Ignored in the case if gaussian=True.
If int provided, determines number of even domain splits
If array of ints, determines number of even domain splits along
each axis
If array of arrays/floats, determines location of splits
antithetic : array_like, optional
List of bool. Represents the axes to mirror using antithetic
variable during MCI.
Returns
-------
out : ndarray
Shapes are related through the identity
k.shape==dist.shape+k.shape
"""
K = np.array(K, dtype=int)
shape = K.shape
dim = len(self)
if dim > 1:
shape = shape[1:]
size = K.size / dim
K = K.reshape(dim, size)
try:
out, G = self.G.run(K, "mom", **kws)
except NotImplementedError:
out = mom(self, K, **kws)
return out.reshape(shape)
def mom(self, K, **kws):
"""Raw statistical moments.
Creates non-centralized raw moments from the random variable. If
analytical options can not be utilized, Monte Carlo integration
will be used.
Parameters
----------
K : array_like
Index of the raw moments. k.shape must be compatible with
distribution shape.
Sampling scheme when performing Monte Carlo
rule : str
rule for estimating the moment if the analytical method
fails.
Key Monte Carlo schemes
--- ------------------------
"H" Halton sequence
"K" Korobov set
"L" Latin hypercube sampling
"M" Hammersley sequence
"R" (Pseudo-)Random sampling
"S" Sobol sequence
Key Quadrature schemes
--- ------------------------
"C" Clenshaw-Curtis
"Q" Gaussian quadrature
"E" Gauss-Legendre
composit : int, array_like optional
If provided, composit quadrature will be used.
Ignored in the case if gaussian=True.
If int provided, determines number of even domain splits
If array of ints, determines number of even domain splits along
each axis
If array of arrays/floats, determines location of splits
antithetic : array_like, optional
List of bool. Represents the axes to mirror using antithetic
variable during MCI.
Returns
-------
out : ndarray
Shapes are related through the identity
k.shape==dist.shape+k.shape
"""
K = np.array(K, dtype=int)
shape = K.shape
dim = len(self)
if dim>1:
shape = shape[1:]
size = K.size/dim
K = K.reshape(dim, size)
try:
out, G = self.G.run(K, "mom", **kws)
except NotImplementedError:
out = mom(self, K, **kws)
return out.reshape(shape)