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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hedman, Marcel, Ivanova, Desi R., Guan, Cong, Rainforth, Tom
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912858688716800
author Hedman, Marcel
Ivanova, Desi R.
Guan, Cong
Rainforth, Tom
author_facet Hedman, Marcel
Ivanova, Desi R.
Guan, Cong
Rainforth, Tom
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2507_14057
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Step-DAD: Semi-Amortized Policy-Based Bayesian Experimental Design
Hedman, Marcel
Ivanova, Desi R.
Guan, Cong
Rainforth, Tom
Machine Learning
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.
title Step-DAD: Semi-Amortized Policy-Based Bayesian Experimental Design
topic Machine Learning
url https://arxiv.org/abs/2507.14057