def make_graph(self) -> None:
# Distribution parameters may be nodes which have named node-inputs
# specified in the point. Need to find the node-inputs, their
# parents and children to replace them.
symbolic_params = self.symbolic_params
self.leaf_nodes = {}
self.named_nodes_parents = {}
self.named_nodes_children = {}
for _, param in symbolic_params:
if hasattr(param, "name"):
# Get the named nodes under the `param` node
nn, nnp, nnc = get_named_nodes_and_relations(param)
self.leaf_nodes.update(nn)
# Update the discovered parental relationships
for k in nnp.keys():
if k not in self.named_nodes_parents.keys():
self.named_nodes_parents[k] = nnp[k]
else:
self.named_nodes_parents[k].update(nnp[k])
# Update the discovered child relationships
for k in nnc.keys():
if k not in self.named_nodes_children.keys():
self.named_nodes_children[k] = nnc[k]
else:
self.named_nodes_children[k].update(nnc[k])
def step(model):
point = {}
for var in model.basic_RVs:
var_name = var.name
if hasattr(var, 'distribution'):
if is_transformed_name(var_name):
val = var.distribution.dist.random(
point=point, size=size)
var_name_untranf = get_untransformed_name(var_name)
point[var_name_untranf] = val
val = var.distribution.transform_used.forward_val(val)
else:
val = var.distribution.random(point=point, size=size)
else:
nn, _, _ = get_named_nodes_and_relations(var)
val = var.eval({model.named_vars[v]: point[v]
for v in nn})
point[var_name] = val
return point
def sample_prior(samples=500, model=None, vars=None, size=None,
random_seed=None, progressbar=True):
"""Generate samples from the prior of a model.
Parameters
----------
samples : int
Number of samples from the prior to generate. Defaults to 500.
model : Model (optional if in `with` context)
vars : iterable
Variables for which to compute the posterior predictive samples.
Defaults to `model.named_vars`.
size : int
The number of random draws from the distribution specified by the
parameters in each sample of the trace.
random_seed : int
Seed for the random number generator.
progressbar : bool
Whether or not to display a progress bar in the command line.
Returns
-------
dict
Dictionary with the variables as keys. The values are arrays of prior samples.
"""
model = modelcontext(model)
if vars is None:
vars = set(model.named_vars.keys())
if random_seed is not None:
np.random.seed(random_seed)
if progressbar:
indices = tqdm(range(samples))
def step(model):
point = {}
for var in model.basic_RVs:
var_name = var.name
if hasattr(var, 'distribution'):
if is_transformed_name(var_name):
val = var.distribution.dist.random(
point=point, size=size)
var_name_untranf = get_untransformed_name(var_name)
point[var_name_untranf] = val
val = var.distribution.transform_used.forward_val(val)
else:
val = var.distribution.random(point=point, size=size)
else:
nn, _, _ = get_named_nodes_and_relations(var)
val = var.eval({model.named_vars[v]: point[v]
for v in nn})
point[var_name] = val
return point
try:
prior = {var: [] for var in vars}
for _ in indices:
point = step(model)
for var_name, var in model.named_vars.items():
if var_name in vars:
if var_name in point:
val = point[var_name]
else:
nn, _, _ = get_named_nodes_and_relations(var)
val = var.eval({model.named_vars[v]: point[v]
for v in nn})
prior[var_name].append(val)
except KeyboardInterrupt:
pass
finally:
if progressbar:
indices.close()
return {k: np.asarray(v) for k, v in prior.items()}