NNest

Neural network accelerated nested and MCMC sampling. The target distribution is first transformed into a diagonal, unit variance Gaussian by a series of non-linear, invertible, and non-volume preserving flows. Efficient MCMC proposals can then be made in this simpler latent space.

Installation

NNest can be installed via pip

pip install nnest

Alternatively the latest version can be obtained by

git clone https://github.com/adammoss/nnest
cd nnest
python setup.py install

Nested Sampling

Nested sampling examples can be found in the examples/nested directory, and can be run with e.g.

python run.py --x_dim 2 --likelihood rosenbrock

There is also an example notebook. Runs can be analysed by

python analyse.py

MCMC Sampling

MCMC sampling (work in progress) examples can be found in the examples/mcmc directory, and can be run with e.g.

python run.py --x_dim 2 --likelihood rosenbrock

There is also an example notebook.

Attribution

Please cite Moss (2019) if you find the package useful in your research.