surrogate-assisted-parallel-tempering

Surrogate-Assisted Parallel Tempering for Bayesian Neural Learning

code

We have two major versions

1. surrogate parallel tempering using random-walk proposal distribution: [surrogate_pt_classifier_rw_common_interval.py] to run with [run_probability_commoninterval.sh]
2. surrogate parallel tempering using Langevin gradient proposal distribution: [surrogate_pt_classifier_langevingrad.py] to run with [run_langevin.sh]

paper online: Surrogate-assisted parallel tempering for Bayesian neural learning

Prerequisites

The framework is built using:

• Paralel tempering neural networks with paralel processing
• Neurocomputing paper by R. Chandra et. al with Arxiv open access

Installing

you need to install Tensorflow and scikitlearn for surrogate training.

Running the tests

Datasets for the experiments are given: data

Experiments

Example results and sh script for running experiment is given here: sample experiment results

Versioning

• TBA

Authors

• R. Chandra, K Jain, A. Kapoor Surrogate-assisted parallel tempering for Bayesian neural learning

License

• This project is licensed under the MIT License - see the Open Source Licence file for details

Acknowledgments

• R. Dietmar Muller and Danial Azam, University of Sydney

Contact

• Dr. Rohitash Chandra, University of New South Wales (c.rohitash at gmail.com or rohitash.chandra at unsw.edu.au)