Kernel Learning for Sample Constrained Black-Box Optimization
This addresses the sample efficiency challenge in black-box optimization for applications like personalized devices and AI-generated content, representing a novel method for a known bottleneck.
The paper tackles the problem of optimizing unknown functions with limited samples by learning the kernel of a Gaussian Process, resulting in a method that outperforms state-of-the-art approaches by achieving optimal estimates with considerably lower sample budgets, as demonstrated in synthetic benchmarks and real applications like hearing aid personalization and generative model convergence.
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.