This repository contains Matlab (R2021a) codes to reproduce the experiments in "Robust Generalised Bayesian Inference for Intractable Likelihoods".

The main function is KSD_Bayes, which performs conjugate inference for exponential family models using KSD-Bayes.

There are four main applications considered in the paper:

1. A Gaussian location model,
2. a two-dimensional scale parameter estimation problem due to Liu et al,
3. density estimation using a kernel exponential family model,
4. an exponential graphical model, and

Results for each application can be reproduced by running the files reproduce_[X].m, where [X] = Gauss, Liu, KEF, EGM.

In addition, there is one more supplementary application:

5. A Ising model (this application is implemented by Pytorch).

A Result for this application can be reproduced by running *.py files with a setting contain in *.sh files.

The "utilities" folder contains routines for regularised covariance estimation, from the "RegularizedSCM" toolbox. See:

Esa Ollila and Elias Raninen, "Matlab RegularizedSCM Toolbox Version 1.11 Available online: http://users.spa.aalto.fi/esollila/regscm/, August 2021.

Esa Ollila and Elias Raninen, "Optimal shrinkage covariance matrix estimation under random sampling from elliptical distributions," arXiv:1808.10188 [stat.ME].