pyGibbsLDA

Python Implementation of Collapsed Gibbs Sampling for Latent Dirichlet Allocation (LDA)

Develop environment

• Language: Python3
• Prerequisite libraries: Scipy, Numpy, matplotlib

Input data format

DocumentID \t WordID \t Count \n

Module usage example

>>> import GibbsLDA
>>> sampler = GibbsLDA.Sampler("/home2/TwitterData.dat", 100)
>>> likelihood = sampler.run(500, 300, 2)

• 100: # of topics
• 500: # of Gibbs samples
• 300: burn-in point
• 2: sampling interval

Reference

• http://www.cs.columbia.edu/~blei/papers/BleiNgJordan2003.pdf
• http://parkcu.com/blog/gibbs-sampling-for-topic-models