About VBLab
VBLab is a probabilistic programming software package, currently available in Matlab, allowing automatic variational Bayesian (VB) inference on many pre-defined common statistical models and also user-defined models.
Key features:
- Providing various Fixed From VB (FFVB) methods and works efficiently for high dimensional and complex posterior distributions.
- Users are not required to know the technicality behind the VB techniques provided; all they need to do is to supply their statistical models, which can be specified flexibly in various ways.
Get started now VBLab on GitHub
Getting started
Install VBLab package
- Download or clone the VBLab package on VBLab Github Page
- Add the VBLab package, with all subfolders, to the Matlab search path. See How to add or remove folders to Matlab search path
How to start
- Read the VB tutorial paper for the theoretical explanation of the VB methods supported by the VBLab package. See also shorter version of the VB tutorial on this site.
- Run examples showing how to use various VB methods to fit different VBLab and user-defined models. See detail explanation of the examples in the the VB tutorial paper or in the Example section on this site.
- Check API reference for supported VB techniques, statistical models and how to define custom models for users’ applications.
Authors
- Trong-Nghia Nguyen, PhD candidate, The University of Sydney Business School. (Google scholar, Research gate, LinkedIn)
- Minh-Ngoc Tran, Associate Professor, The University of Sydney Business School. (Google scholar, Research gate, Home Page)
- Viet-Hung Dao, PhD Candidate, The University of New South Wales Business School. (Google scholar, Research gate, Home Page)
Citing VBLab
If you use VBLab in a scientific publication, we would appreciate citations to the following paper:
M.-N. Tran, T.-N. Nguyen and V.-H. Dao (2021). A practical tutorial on Variational Bayes. arXiv2103.01327.
Or bibtex entry :
@misc{tran:2021,
title={A practical tutorial on Variational Bayes},
author={Minh-Ngoc Tran and Trong-Nghia Nguyen and Viet-Hung Dao},
year={2021},
eprint={2103.01327},
archivePrefix={arXiv},
primaryClass={stat.CO}
}