def __init__(self,
name,
learnable,
ranges,
is_observed=False,
**kwargs): #TODO: code duplication here
self.name = name
self._evaluated = False
self._observed = is_observed
self._observed_value = None
self._current_value = None
self.construct_deterministic_parents(learnable, ranges, kwargs)
self.parents = join_sets_list(
[var2link(x).vars for x in kwargs.values()])
self.ancestors = join_sets_list(
[self.parents] + [parent.ancestors for parent in self.parents])
self.link = LinkConstructor(**kwargs)
self.samples = None
self.ranges = {}
self.dataset = None
self.has_random_dataset = False
self.has_observed_value = False
self.is_normalized = True
self.partial_links = {
name: var2link(link)
for name, link in kwargs.items()
}
def __init__(self,
name,
learnable,
ranges,
is_observed=False,
has_bias=False,
**kwargs):
self._input = kwargs.copy()
self.name = name
self.learnable = learnable
self._bias = has_bias
self._evaluated = False
self._observed = is_observed
self._observed_value = None
self._current_value = None
if self._check_for_stochastic_process_arguments():
return
if has_bias:
self.construct_biases(learnable, ranges, kwargs)
self.construct_deterministic_parents(learnable, ranges, kwargs)
self.parents = join_sets_list(
[var2link(x).vars for x in kwargs.values()])
self.link = Link(**kwargs)
self.ancestors = join_sets_list(
[self.parents] + [parent.ancestors for parent in self.parents])
self.samples = None
self.ranges = {}
self.dataset = None
self.has_random_dataset = False
self.has_observed_value = False
self.is_normalized = True
self.partial_links = {
name: var2link(link)
for name, link in kwargs.items()
}
def __init__(self, name, learnable, ranges, is_observed=False, has_bias=False, is_policy=False, is_reward=False, **kwargs):
self._input = kwargs
self.name = name
self._evaluated = False
self._observed = is_observed
self._observed_value = None
self._current_value = None
#if self._check_for_stochastic_process_arguments():
# return
if has_bias:
self.construct_biases(learnable, ranges, kwargs)
self.construct_deterministic_parents(learnable, ranges, kwargs)
self.parents = join_sets_list([var2link(x).vars for x in kwargs.values()])
self.link = Link(**kwargs)
self.ancestors = join_sets_list([self.parents] + [parent.ancestors for parent in self.parents])
self.samples = None
self.ranges = {}
self.dataset = None
self.has_random_dataset = False
self.has_observed_value = False
self.is_normalized = True
self.silenced = False
self.partial_links = {name: var2link(link) for name, link in kwargs.items()}
# Reinforcement learning descriptors
if not (is_policy and is_reward):
self._policy = is_policy
self._reward = is_reward
else:
raise ValueError("A variable cannot be both a RL policy and a RL reward")
def __init__(self, name, learnable, ranges, is_observed=False, **kwargs):
class VarLink(chainer.ChainList):
def __init__(self):
self.kwargs = kwargs
links = [
link for partial_link in kwargs.values()
for link in var2link(partial_link).links
]
super().__init__(*links)
def __call__(self, values):
return {
k: var2link(x).fn(values)
for k, x in self.kwargs.items()
}
self.name = name
self._evaluated = False
self._observed = is_observed
self._observed_value = None
self._current_value = None
self.construct_deterministic_parents(learnable, ranges, kwargs)
self.parents = join_sets_list(
[var2link(x).vars for x in kwargs.values()])
self.link = VarLink()
self.samples = []
self.ranges = {}
self.dataset = None
self.has_random_dataset = False
self.has_observed_value = False
def var2link(var):
if isinstance(var, Variable):
vars = {var}
fn = lambda values: values[var]
elif isinstance(var, (numbers.Number, np.ndarray, torch.Tensor)):
vars = set()
fn = lambda values: var
elif isinstance(var, (tuple, list)) and all([isinstance(v, (Variable, PartialLink)) for v in var]):
vars = join_sets_list([{v} if isinstance(v, Variable) else v.vars for v in var])
fn = lambda values: tuple([values[v] if isinstance(v, Variable) else v.fn(values) for v in var])
else:
return var
return PartialLink(vars=vars, fn=fn, links=set(), string=str(var))
def _flatten(self):
variables = list(
join_sets_list(
[var.ancestors.union({var}) for var in self.variables]))
return sorted(variables, key=lambda v: v.name)
