Projective Preferential Bayesian Optimization

Petrus Mikkola, Milica Todorović, Jari Järvi, Patrick Rinke, Samuel Kaski
2/8/2020
stat.MLcs.LG

Official code: AaltoPML/PPBO (20 stars, MIT, last commit 2023-07-06).

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.

View on arXivView PDF

Code Implementations(1)

AaltoPML/PPBOOfficial100%
MIT

Projective Preferential Bayesian Optimization

200Shell, PythonJun 29, 20203 years agoMIT

Similar Papers

Experience in Engineering Complex Systems: Active Preference Learning with Multiple Outcomes and Certainty Levels

85% match2023

Local Preferential Bayesian Optimization

84% match2026

Polynomial-Model-Based Optimization for Blackbox Objectives

83% match2023

Generative Refinement for Low-Budget Black-Box Optimization

82% match2026

Direct Preference-Based Evolutionary Multi-Objective Optimization with Dueling Bandit

82% match2023

Bayesian Optimization of Combinatorial Structures

82% match1 repo1052018

Cite this paper

@article{mikkola2020projective,
  title  = {Projective Preferential Bayesian Optimization},
  author = {Petrus Mikkola and Milica Todorović and Jari Järvi and Patrick Rinke and Samuel Kaski},
  year   = {2020},
  eprint = {2002.03113},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2002.03113}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2002.03113.svg)](https://codeofpaper.com/paper/2002.03113)

Discussion