gistfile1.py

# Run models to find natural number of topics
kl = []
l = np.array([sum(cnt for _, cnt in doc) for doc in my_corpus])
num = range(1,150,1)
for i in num:
    lda = models.ldamodel.LdaModel(corpus=my_corpus,
        id2word=dictionary,num_topics=i)
    #Topic-word matrix
    m1 = lda.expElogbeta
    U,cm1,V = np.linalg.svd(m1)
    #Document-topic matrix
    lda_topics = lda[my_corpus]
    m2 = matutils.corpus2dense(lda_topics, lda.num_topics).transpose()
    cm2 = l.dot(m2)
    # cm2 = cm2 + 0.0001
    cm2norm = np.linalg.norm(l)
    cm2 = cm2/cm2norm
    div = sym_kl(cm1,cm2)
    kl.append(div)

# Plot kl divergence against number of topics
plt.plot(kl)
plt.ylabel('Symmetric KL Divergence')
plt.xlabel('Number of Topics')
plt.savefig('kldiv.png', bbox_inches='tight')