PYSGMCMC is a Python framework for Bayesian Deep Learning that focuses on Stochastic Gradient Markov Chain Monte Carlo methods.
- Complex samplers as black boxes, computing the next sample with corresponding costs of any MCMC sampler is as easy as:
sample, cost = next(sampler)- Based on tensorflow that provides:
- efficient numerical computation via data flow graphs
- flexible computation environments (CPU/GPU support, desktop/server/mobile device support)
- Linear algebra operations
The quick way:
pip3 install git+https://github.com/MFreidank/pysgmcmc
Our documentation can be found at http://pysgmcmc.readthedocs.io/en/latest/.