Lazier ABC

TL;DR

Lazy ABC reduces computational costs by early stopping of unpromising simulations using a random decision process, maintaining the same target distribution as standard ABC, and is extended to non-uniform kernels for easier tuning.

Contribution

This paper extends the lazy ABC algorithm to non-uniform kernels and demonstrates its effectiveness in reducing computational costs while preserving accuracy.

Findings

Lazy ABC significantly decreases simulation time.

Extension to non-uniform kernels simplifies tuning.

Target distribution remains unchanged with lazy stopping.

Abstract

ABC algorithms involve a large number of simulations from the model of interest, which can be very computationally costly. This paper summarises the lazy ABC algorithm of Prangle (2015), which reduces the computational demand by abandoning many unpromising simulations before completion. By using a random stopping decision and reweighting the output sample appropriately, the target distribution is the same as for standard ABC. Lazy ABC is also extended here to the case of non-uniform ABC kernels, which is shown to simplify the process of tuning the algorithm effectively.