Kernel Learning for Sample Constrained Black-Box Optimization

TL;DR

This paper introduces KOBO, a kernel learning method for black-box optimization that reduces sample costs by learning kernels in a latent space, improving performance on benchmarks and real-world tasks.

Contribution

It proposes a novel kernel learning approach using a variational autoencoder to optimize Gaussian Process kernels, enhancing sample efficiency in black-box optimization.

Findings

KOBO outperforms existing methods on synthetic benchmarks.

KOBO achieves effective personalization with fewer queries.

Real-world applications demonstrate practical benefits.

Abstract

Black box optimization (BBO) focuses on optimizing unknown functions in high-dimensional spaces. In many applications, sampling the unknown function is expensive, imposing a tight sample budget. Ongoing work is making progress on reducing the sample budget by learning the shape/structure of the function, known as kernel learning. We propose a new method to learn the kernel of a Gaussian Process. Our idea is to create a continuous kernel space in the latent space of a variational autoencoder, and run an auxiliary optimization to identify the best kernel. Results show that the proposed method, Kernel Optimized Blackbox Optimization (KOBO), outperforms state of the art by estimating the optimal at considerably lower sample budgets. Results hold not only across synthetic benchmark functions but also in real applications. We show that a hearing aid may be personalized with fewer audio queries to the user, or a generative model could converge to desirable images from limited user ratings.