Projective Preferential Bayesian Optimization
This addresses the challenge of interactive preference learning in high-dimensional settings, such as molecular adsorption, offering a novel query method that is more feasible than pairwise comparisons.
The paper tackles the problem of learning user preferences in high-dimensional spaces where direct function evaluation is impossible, proposing projective preferential Bayesian optimization that uses queries along projections, and demonstrates it can find global minima in high dimensions, which existing preferential methods fail to do.
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.