Latent Dirichlet allocation

The lda.py module implements:

1. Standard (ie., non-Bayesian) latent Dirichlet allocation with MLE parameter estimates via variational EM (Blei, Ng, & Jordan, 2003).
2. Fully-Bayesian (ie., smoothed) latent Dirichlet allocation with MAP parameter estimates via collapsed Gibbs sampling (Griffiths & Steyvers, 2004).

Plots

Unsmoothed

### Smoothed TODO