def setUp(self):
num_topics = 3
vocab_size = 14
alpha = np.array([0.1, 0.2, 0.3])
eta = np.array(range(1, vocab_size+1)) / 100.0
self.model = lda.LdaModel(corpus, num_topics, alpha, eta, 0, 0)
self.model.stats = stats
def main():
# Create an image with one row per iteration to visualize results
topic_img = np.zeros((NUM_ITER + 2, NUM_TOPICS * VOCAB_SIZE))
print 'Generating corpus'
corpus, beta = generate_corpus()
# Add real topics to the image
add_to_img(topic_img, NUM_ITER + 1, beta, beta)
alpha_guess = 0.5
eta_guess = 0.5
print 'Initializing model'
model = lda.LdaModel(corpus, NUM_TOPICS, alpha_guess, eta_guess,
GIBBS_BURN, GIBBS_LAG)
add_to_img(topic_img, 0, beta, model.beta())
# Do E-M iterations
last_lik = model.log_likelihood_wz()
likelihoods = [(0, last_lik)]
for i in range(NUM_ITER):
print 'Iteration %d' % i
model.e_step()
lik = model.log_likelihood_wz()
print ' E-step ML: %f (%f)' % (lik, lik - last_lik)
last_lik = lik
model._m_alpha()
lik = model.log_likelihood_wz()
print ' M-step(alpha) ML: %f (%f)' % (lik, lik - last_lik)
last_lik = lik
model._m_eta()
add_to_img(topic_img, i + 1, beta, model.beta())
lik = model.log_likelihood_wz()
print ' M-step(eta) ML: %f (%f)' % (lik, lik - last_lik)
last_lik = lik
likelihoods.append((i + 1, lik))
save_likelihoods(likelihoods)
save_topics(topic_img)
def main():
print 'Generating corpus'
corpus = generate_corpus()
print 'Initializing Model'
model = lda.LdaModel(corpus, num_topics, alpha, 0, 0)
print 'Sampling'
for i in range(ITERATIONS):
model._gibbs_sample(model.stats)
print 'Iteration %d complete' % (i+1)
save_beta(model.beta(), i)
def setUp(self):
num_topics = 3
vocab_size = 14
self.model = lda.LdaModel(corpus, num_topics, stub_alpha, stub_eta)
self.model.stats = stats
def test_init_default(self):
model = lda.LdaModel(corpus, self.num_topics)
alpha = np.ones(self.num_topics) * 0.1
eta = np.ones(self.vocab_size) * 0.1
nptest.assert_array_equal(model.alpha, alpha)
nptest.assert_array_equal(model.eta, eta)
def test_init_vector(self):
alpha = np.ones(self.num_topics) * 0.2
eta = np.ones(self.vocab_size) * 0.3
model = lda.LdaModel(corpus, self.num_topics, alpha, eta)
nptest.assert_array_equal(model.alpha, alpha)
nptest.assert_array_equal(model.eta, eta)
# -*- coding:utf-8 -*-
import lda
import simplejson as sj
import os
import time
ISOTIMEFORMAT = '%Y-%m-%d %X'
print 'start at:' + time.strftime(ISOTIMEFORMAT, time.localtime())
p = lda.LdaModel()
p.readFile('YOUR DATA DIRECTION')
print 'file reading finish...'
p.DocumentInitiation()
print 'initiation finish...'
p.lda_inference()
print 'lda inference...'
print 'end at:' + time.strftime(ISOTIMEFORMAT, time.localtime())