Skip to content

Latest commit

 

History

History
22 lines (14 loc) · 1.23 KB

README.md

File metadata and controls

22 lines (14 loc) · 1.23 KB

Quasi-Newton Coresets: Fast Bayesian coresets with theoretical guarantees

This repository provides a python package to construct Bayesian coresets. The code is based on that found here. It contains code to run the experiments in Fast Bayesian Coresets via Subsampling and Quasi-Newton Refinement.

Installation and Dependencies

To install with pip, download the repository and run pip3 install . --user in the repository's root folder.

The experiments depend on NumPy, SciPy, SciKit Learn, PyStan, pandas and Bokeh.

Instructions

In order to run the synthetic Gaussian experiment, the data needs to be generated by running the get_synthetic_gaussian_data.py file within the data folder. This is because the dataset is too large to fit within the size limit.

Having done this, all experiments can be run by running run_experiment.sh under each directory. To plot the results, run make_plots.sh under each directory.