class TopN(MCMCSummary, SampleFilter):
"""
This "summary" uses a FiniteBestSet to store the top N
It replaces where we previously used a FiniteBestSet
"""
def __init__(self, N=1000, key='posterior_score', thin=1):
self.__dict__.update(locals())
self.fbs = FiniteBestSet(N=N, key=key)
self.count = 0
def add(self, h):
self.count += 1
if self.count % self.thin == 0:
self.fbs.add(h)
def get_all(self, sorted=False):
for h in self.fbs.get_all(sorted=sorted):
yield h
def __iter__(self):
for h in self.fbs.get_all():
yield h
def Z(self):
"""
Normalizer of everything
"""
return logsumexp([h.posterior_score for h in self.get_all(sorted=False)])
def normalize(self, d):
"""
Change the posterior_score on hypotheses so they sum to 1
"""
Z = self.Z()
for h in self.fbs.get_all(sorted=False):
h.posterior_score = h.posterior_score - Z
def display(self):
for h in self.get_all():
print h.posterior_score, h.prior, h.likelihood, qq(h)
def update(self, y):
if isinstance(y, TopN):
assert y.key == self.key
for h in y.fbs.get_all(sorted=False):
self.add(h)
elif isinstance(y, collections.Iterable):
for yi in y:
self.update(yi)
elif isinstance(y, FiniteBestSet):
self.fbs.merge(y)
else:
raise NotImplementedError