Knowledge-based Selection of Gaussian Process Surrogates

0509320 - ÚI 2020 RIV DE eng C - Konferenční příspěvek (zahraniční konf.)
Pitra, Zbyněk - Bajer, Lukáš - Holeňa, Martin
Knowledge-based Selection of Gaussian Process Surrogates.
IAL ECML PKDD 2019: Workshop & Tutorial on Interactive Adaptive Learning. Proceedings. Aachen: Technical University & CreateSpace Independent Publishing Platform, 2019 - (Kottke, D.; Lemaire, D.; Calma, A.; Krempl, G.; Holzinger, A.), s. 48-63. CEUR Workshop Proceedings, 2444. ISSN 1613-0073.
[ECML PKDD 2019: The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases. Würzburg (DE), 16.09.2019-20.09.2019]
Grant CEP: GA ČR GA17-01251S; GA ČR(CZ) GA18-18080S
Grant ostatní: ČVUT(CZ) SGS17/193/OHK4/3T/14; GA MŠk(CZ) LM2015042
Institucionální podpora: RVO:67985807
Klíčová slova: Benchmarking * Black-box optimization * Gaussian process * Landscape analysis
Obor OECD: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
http://ceur-ws.org/Vol-2444/ialatecml_paper4.pdf

Many real-world problems belong to the area of continuous black-box optimization. If the black-box function is also cost-aware, regression surrogate models are often utilized by optimization algorithms to save evaluations of the original cost-aware function. Choosing a suitable surrogate model or a suitable setting of its hyperparameters is a complex selection problem, where research into reusing knowledge represented by features of black-box function landscape is only starting. In this paper, we report the research into surrogate model selection, where knowledge from the previous experience with using the model is utilized to design a metalearing system. As a proof of concept, we provide a study investigating the influence of landscape features on the performance of various Gaussian process covariance functions as surrogate models for the state-of-the-art optimization algorithm in the cost-aware continuous black-box optimization.
Trvalý link: http://hdl.handle.net/11104/0300063