Step-DAD: Semi-Amortized Policy-Based Bayesian Experimental Design

TL;DR

This paper introduces Step-DAD, a semi-amortized policy-based Bayesian experimental design method that adapts its design policy during data collection, leading to improved decision-making and robustness.

Contribution

It proposes a novel semi-amortized approach that updates the design policy during experiments, enhancing flexibility and robustness over existing methods.

Findings

Step-DAD outperforms state-of-the-art BED methods in decision quality.

It demonstrates increased robustness across different experimental scenarios.

The approach effectively balances pre-trained policies with adaptive updates.

Abstract

We develop a semi-amortized, policy-based, approach to Bayesian experimental design (BED) called Stepwise Deep Adaptive Design (Step-DAD). Like existing, fully amortized, policy-based BED approaches, Step-DAD trains a design policy upfront before the experiment. However, rather than keeping this policy fixed, Step-DAD periodically updates it as data is gathered, refining it to the particular experimental instance. This test-time adaptation improves both the flexibility and the robustness of the design strategy compared with existing approaches. Empirically, Step-DAD consistently demonstrates superior decision-making and robustness compared with current state-of-the-art BED methods.