- Support parallelized sampler proposed in Distributed Inference for Latent Dirichlet Allocation.
- Implement CGS_p estimator for more precise point estimate of topic-word distribution.
- Implement Labelled LDA
- Able to obtain per-word topic frequency.
The implementaion relies on Eigen for faster array multiplication and pybind11 for simple binding.
You can install the wheel from pypi:
pip install lda11
For x64 architecture, the above wheel is built using AVX. If it is not convenient for you, try e.g.
CFLAGS="-march=native" pip install git+https://github.com/tohtsky/lda11