MO-SMAC: Multiobjective Sequential Model-Based Algorithm Configuration

Evolutionary Computation, vol. 34, pp. 29–52

Abstract

Automated algorithm configuration aims at finding well-performing parameter configurations for a given problem, and it has proven to be effective within many AI domains, including evolutionary computation. Initially, the focus was on excelling in one performance objective, but, in reality, most tasks have a variety of (conflicting) objectives. The surging demand for trustworthy and resource-efficient AI systems makes this multiobjective perspective even more prevalent. We propose a new general-purpose multiobjective automated algorithm configurator by extending the widely-used SMAC framework. Instead of finding a single configuration, we search for a nondominated set that approximates the actual Pareto set. We propose a pure multiobjective Bayesian optimization approach for obtaining promising configurations by using the predicted hypervolume improvement as acquisition function. We also present a novel intensification procedure to efficiently handle the selection of configurations in a multiobjective context. Our approach is empirically validated and compared across various configuration scenarios in four AI domains, demonstrating superiority over baseline methods, competitiveness with MO-ParamILS on individual scenarios, and an overall best performance.

Authors 6

1. Jeroen Rook

   University of Twente

   Affiliation as printed

   Data Management & Biometrics, University of Twente, The Netherlands j.g.rook@utwente.nl

2. Carolin Benjamins

   Leibniz University Hannover

   Affiliation as printed

   Institute of Artificial Intelligence, Leibniz University Hannover, Germany c.benjamins@ai.uni-hannover.de

3. Jakob Bossek

   Paderborn University

   Affiliation as printed

   Machine Learning and Optimisation, Paderborn University, Germany jakob.bossek@uni-paderborn.de

4. Heike Trautmann

   University of Twente · Paderborn University

   Affiliation as printed

   Data Management & Biometrics, University of Twente, The Netherlands

   Machine Learning and Optimisation, Paderborn University, Germany heike.trautmann@uni-paderborn.de

   Data Management & Biometrics, University of Twente, The Netherlands heike.trautmann@uni-paderborn.de

   Machine Learning and Optimisation, Paderborn University, Germany

5. Holger H. Hoos Aachen Artificial Intelligence Methodology

   RWTH Aachen University · Leiden University · University of British Columbia

   Affiliation as printed

   ADA Research Group, Leiden University, The Netherlands

   Artificial Intelligence Methodology, RTWH Aachen University, Germany hh@aim.rwth-aachen.de

   University of British Columbia, Canada

   Artificial Intelligence Methodology, RTWH Aachen University, Germany

   University of British Columbia, Canada hh@aim.rwth-aachen.de

6. Marius Lindauer

   Leibniz University Hannover · L3S Research Center

   Affiliation as printed

   Institute of Artificial Intelligence, L3S Research Center Leibniz University Hannover, Germany m.lindauer@ai.uni-hannover.de