def __init__(self, opts):
if isinstance(opts, dict):
self.options = []
self.weights = dict()
ordered = []
for opt, prob in opts.items():
if not isinstance(prob, (float, int)):
raise RuntimeParseError(
f'discrete distribution weight {prob}'
' is not a constant number')
if prob < 0:
raise RuntimeParseError(
f'discrete distribution weight {prob} is negative')
if prob == 0:
continue
opt = toDistribution(opt)
self.options.append(opt)
self.weights[opt] = prob
ordered.append(prob)
self.cumulativeWeights = tuple(itertools.accumulate(ordered))
else:
self.options = tuple(toDistribution(opt) for opt in opts)
self.cumulativeWeights = None
if len(self.options) == 0:
raise RuntimeParseError(
'tried to make discrete distribution over empty domain!')
valueType = type_support.unifyingType(self.options)
super().__init__(*self.options, valueType=valueType)
def __init__(self, opts): self.options = opts valueType = type_support.unifyingType(self.options) super().__init__(*self.options, valueType=valueType)
def __init__(self, index, options): self.index = index self.options = tuple(toDistribution(opt) for opt in options) assert len(self.options) > 0 valueType = type_support.unifyingType(self.options) super().__init__(index, *self.options, valueType=valueType)