Bayesin Experimental Design

Explore the Bayesin Optimal Experimental Design under model misspecification.

Bayesian Optimal Experimental Design (BOED) is crucial for adaptive experimentation, where the goal is to maximise information gain about parameters of interest through sequential decision-making. BOED’s significance lies in its ability to optimise experimental processes, reducing costs and time. However, traditional BOED methods assume the model is well-specified, which can lead to suboptimal designs and poor data quality when the model is misspecified in real-world scenarios. Addressing this challenge is essential for making BOED robust and widely applicable. This research aims to develop methodologies to mitigate these challenge.