Optimal Bayesian design for model discrimination via classification

TL;DR

This paper introduces a supervised classification-based method for Bayesian optimal model discrimination design, significantly reducing computational costs and enabling effective model comparison even with intractable likelihoods.

Contribution

It proposes a novel classification approach that simplifies Bayesian design for model discrimination, especially useful with complex or intractable likelihood functions.

Findings

Requires fewer simulations than previous methods

Easily assesses design performance via misclassification error

Effective with models having intractable likelihoods

Abstract

Performing optimal Bayesian design for discriminating between competing models is computationally intensive as it involves estimating posterior model probabilities for thousands of simulated datasets. This issue is compounded further when the likelihood functions for the rival models are computationally expensive. A new approach using supervised classification methods is developed to perform Bayesian optimal model discrimination design. This approach requires considerably fewer simulations from the candidate models than previous approaches using approximate Bayesian computation. Further, it is easy to assess the performance of the optimal design through the misclassification error rate. The approach is particularly useful in the presence of models with intractable likelihoods but can also provide computational advantages when the likelihoods are manageable.