Python implementation of Latent Dirichlet Allocation based on Scalable Tensor Orthogonal Decomposition (STOD) from:
Scalable Moment-Based Inference for Latent Dirichlet Allocation
C. Wang, X. Liu, Y. Song, and J. Han, 2014
and https://github.com/UIUC-data-mining/STOD.
The tests/test_lda.py file demonstrates the use of STOD-based LDA to perform topic analysis on the NYT dataset (included in the tests/assets folder) and the 20 newsgroups dataset (from sklearn).
It also implements the algorithm's evaluation using synthetic data from Want et. al. 2014.