FastLDA -- A minimal and fast implementation of Latent Dirichlet Allocation

Still a work in progress, but currently fully usable!

Prerequisites

On Unix (Linux, OS X)

• A compiler with C++11 support
• CMake >= 2.8.12

On Windows

• Visual Studio 2015 (required for all Python versions, see notes below)
• CMake >= 3.1

Installation

Building from source

1. Fork/clone this repo.

2. In fastlda/ please run

git submodule init
git submodule update
pip install --upgrade .  # recommend doing this in a virtualenv

This should install all the dependencies needed.

3. Run

python src/python/lda_example.py

for an end-to-end LDA run (from training to inference using NIPS dataset from https://archive.ics.uci.edu/ml/datasets/bag+of+words).

Alternatives

• Pypi version soon to come!

Usage

A Python class has been exposed with all the relevant training/inference functions.

For an example, see: this simple example

License

MIT

This code is written in C++ with python modules exposed using pybind11.