We investigate several reinforcement learning algorithms on three Bayesian experimental design problems. Performance is measured by each agent's training time and generalisability to various experimental setups at evaluation time.

Source code for Bayesian Experimental Design for Implicit Models by Mutual Information Neural Estimation, ICML 2020, https://arxiv.org/abs/2002.08129

Code for the paper "Gradient-Based Bayesian Experimental Design for Implicit Models using Mutual Information Lower Bounds" https://arxiv.org/abs/2105.04379

Python code for "Implicit Deep Adaptive Design: Policy-Based Experimental Design without Likelihoods", NeurIPS, 2021, https://proceedings.neurips.cc/paper/2021/hash/d811406316b669ad3d370d78b51b1d2e-Abstract.html

Code for the paper "Sequential Bayesian Experimental Design for Implicit Models via Mutual Information", Bayesian Analysis 2021, https://arxiv.org/abs/2003.09379.

Source code for "Efficient Bayesian Experimental Design for Implicit Models", AISTATS 2019, https://arxiv.org/abs/1810.09912

In this work, we develop a new framework for designing experiments that are robust to model misspecification through generalised Bayesian inference. This repository contains the files needed to perform Generalised Bayesian Optimal Experimental Design (GBOED) on several experimental design problems.