passglm: a package for creating and evaluating PASS-GLM models

The passglm package was used produce the experiments for:

Jonathan H. Huggins, Ryan P. Adams, Tamara Broderick. PASS-GLM: polynomial approximate sufficient statistics for scalable Bayesian GLM inference . In Proc. of the 31st Annual Conference on Neural Information Processing Systems (NIPS), 2017.

The package includes functionality to load data, construct PASS-GLM approximations for logistic regression, run an adaptive Metropolis-Hastings sampler, and compare performance of PASS-GLM inferences to those obtained with other methods. Support for streaming and distributed inference is included.

Compilation and testing

To compile and test the package (for development purposes):

python build_ext --inplace  # compile cython code in place
nosetests tests/                     # run tests, which takes a minute or so

To install:

pip install .


For example usages, see the scripts/ directory.