def generate(topics, words, words_per_doc):
num_docs = len(topics)
word_cdfs = [util.get_cdf(topic) for topic in words]
docs = []
doc_topics = []
for i in range(num_docs):
if i % 100 == 0:
print "reached document", i
num_words = util.poisson(words_per_doc)
topic_dist = topics[i]
topic_cdf = util.get_cdf(topic_dist)
doc = []
word_topics = []
for word in range(num_words):
topic = util.sample(topic_cdf)
doc.append(util.sample(word_cdfs[topic]))
word_topics.append(topic)
docs.append(doc)
doc_topics.append(word_topics)
return docs, doc_topics
def __init__(self, params):
# The word distribution of this node's topic.
self.word_dist = dirichlet(params["topic_to_word_param"])
self.word_cdf = util.get_cdf(self.word_dist)
# The number of documents that pass through this node.
self.num_documents = 0
# Those children of this node which have looked below this level.
# Documents that reached this node but never looked below aren't
# represented here; this is okay because the Chinese Restaurant
# Process is exchangeable (doesn't depend on order).
self.children = []
# The number of documents which looked below this level. This
# should always be equal to sum(c.num_documents for c in
# self.children).
self.num_documents_in_children = 0
def new_customer(self):
self.customers += 1
self.update_probabilities()
assert(abs(1 - sum(self.probabilities)) < 1e-10)
cdf = util.get_cdf(self.probabilities)
table = util.sample(cdf)
print "new customer has arrived!", self.customers
#print "customers are sitting at tables", self.sparse_seats()
#print "customer will sit with probabilities:", self.probabilities
print "customer has chosen to sit at table", table
if table not in self.seats:
self.tables.append(1)
else:
self.tables[table] += 1
self.seats.append(table)
