Sequential Bayesian Experimental Design for Implicit Models via Mutual Information

Bayesian experimental design (BED) is a framework that uses statistical models and decision making under uncertainty to optimise the cost performance of scientific experiment. Sequential BED, as opposed static considers scenario where we can sequentially update our beliefs about model parameters through data gathered in A class particular interest for natural medical sciences are implicit models, generating distribution intractable, but sampling from it possible. Even though there has been lot work on BED past few years, notoriously difficult problem sequential barely touched upon. We address this gap literature by devising novel parameter estimation Mutual Information (MI) between simulated utility function find optimal designs, which not done before models. Our approach likelihood-free inference ratio simultaneously estimate posterior distributions MI. During procedure utilise optimisation help us MI utility. efficient various tested, yielding accurate estimates after only iterations.