Abstract

Bayesian optimization is an effective method for finding extrema of a black-box function. We propose a new type of Bayesian optimization for learning user preferences in high-dimensional spaces. The central assumption is that the underlying objective function cannot be evaluated directly, but instead a minimizer along a projection can be queried, which we call a projective preferential query. The form of the query allows for feedback that is natural for a human to give, and which enables interaction. This is demonstrated in a user experiment in which the user feedback comes in the form of optimal position and orientation of a molecule adsorbing to a surface. We demonstrate that our framework is able to find a global minimum of a high-dimensional black-box function, which is an infeasible task for existing preferential Bayesian optimization frameworks that are based on pairwise comparisons.
Original languageEnglish
Number of pages9
Publication statusAccepted/In press - 1 Jun 2020
MoE publication typeNot Eligible
EventInternational Conference on Machine Learning - Vienna, Austria
Duration: 12 Jul 202018 Jul 2020
Conference number: 37

Conference

ConferenceInternational Conference on Machine Learning
Abbreviated titleICML
CountryAustria
CityVienna
Period12/07/202018/07/2020

Keywords

  • human-in-the-loop machine learning
  • gaussian process
  • preference learning
  • Bayesian optimization
  • Bayesian methods
  • machine learning
  • expert elicitation

Fingerprint Dive into the research topics of 'Projective Preferential Bayesian Optimization'. Together they form a unique fingerprint.

  • Cite this

    Mikkola, P., Todorovic, M., Järvi, J., Rinke, P., & Kaski, S. (Accepted/In press). Projective Preferential Bayesian Optimization. Paper presented at International Conference on Machine Learning, Vienna, Austria. https://arxiv.org/abs/2002.03113