Angus Phillips

Bayesian Experimental Design via Score Matching

Angus Phillips, Gavin Kerrigan, Tom Rainforth

Policy-based approaches to Bayesian experimental design (BED) allow the learning of deep policy networks that adaptively make intelligent design decisions based on previously collected data. However, the training of such policies is often held back by a fundamental challenge: the…Read more