Pareto-Optimal Anytime Algorithms via Bayesian Racing

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wurth, Jonathan, Stegherr, Helena, Kemper, Neele, Heider, Michael, Hähner, Jörg
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911499123949568
author Wurth, Jonathan
Stegherr, Helena
Kemper, Neele
Heider, Michael
Hähner, Jörg
author_facet Wurth, Jonathan
Stegherr, Helena
Kemper, Neele
Heider, Michael
Hähner, Jörg
contents Selecting an optimization algorithm requires comparing candidates across problem instances, but the computational budget for deployment is often unknown at benchmarking time. Current methods either collapse anytime performance into a scalar, require manual interpretation of plots, or produce conclusions that change when algorithms are added or removed. Moreover, methods based on raw objective values require normalization, which needs bounds or optima that are often unavailable and breaks coherent aggregation across instances. We propose a framework that formulates anytime algorithm comparison as Pareto optimization over time: an algorithm is non-dominated if no competitor beats it at every timepoint. By using rankings rather than objective values, our approach requires no bounds, no normalization, and aggregates coherently across arbitrary instance distributions without requiring known optima. We introduce PolarBear (Pareto-optimal anytime algorithms via Bayesian racing), a procedure that identifies the anytime Pareto set through adaptive sampling with calibrated uncertainty. Bayesian inference over a temporal Plackett-Luce ranking model provides posterior beliefs about pairwise dominance, enabling early elimination of confidently dominated algorithms. The output Pareto set together with the posterior supports downstream algorithm selection under arbitrary time preferences and risk profiles without additional experiments.
format Preprint
id arxiv_https___arxiv_org_abs_2603_08493
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Pareto-Optimal Anytime Algorithms via Bayesian Racing
Wurth, Jonathan
Stegherr, Helena
Kemper, Neele
Heider, Michael
Hähner, Jörg
Neural and Evolutionary Computing
Machine Learning
Selecting an optimization algorithm requires comparing candidates across problem instances, but the computational budget for deployment is often unknown at benchmarking time. Current methods either collapse anytime performance into a scalar, require manual interpretation of plots, or produce conclusions that change when algorithms are added or removed. Moreover, methods based on raw objective values require normalization, which needs bounds or optima that are often unavailable and breaks coherent aggregation across instances. We propose a framework that formulates anytime algorithm comparison as Pareto optimization over time: an algorithm is non-dominated if no competitor beats it at every timepoint. By using rankings rather than objective values, our approach requires no bounds, no normalization, and aggregates coherently across arbitrary instance distributions without requiring known optima. We introduce PolarBear (Pareto-optimal anytime algorithms via Bayesian racing), a procedure that identifies the anytime Pareto set through adaptive sampling with calibrated uncertainty. Bayesian inference over a temporal Plackett-Luce ranking model provides posterior beliefs about pairwise dominance, enabling early elimination of confidently dominated algorithms. The output Pareto set together with the posterior supports downstream algorithm selection under arbitrary time preferences and risk profiles without additional experiments.
title Pareto-Optimal Anytime Algorithms via Bayesian Racing
topic Neural and Evolutionary Computing
Machine Learning
url https://arxiv.org/abs/2603.08493