The article introduces Action-BED, a new framework for Bayesian experimental design that formulates the problem in terms of expected future loss on downstream actions rather than uncertainty reduction. This approach converts traditionally doubly intractable objectives into singly intractable ones that can be jointly optimized using stochastic gradients.
- The method avoids explicit posterior or marginal likelihood estimation by relying only on sampling from the joint model and evaluating the downstream loss function.
- It enables the joint optimization of both the design policy and a downstream action policy through stochastic gradients.
- This formulation allows for easier customization to different downstream tasks and losses compared to existing methods.
This approach allows design policies to be learned more effectively, efficiently, and simply while providing easy customization to various downstream tasks.