Black-Box Gradient Matching for Reliable Offline Black-Box Optimization

Minh Hoang, Azza Fadhel, Aryan Deshwal, Jana Doppa, Trong Nghia Hoang

Primary Area: general machine learning (i.e., none of the above)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: Offline Optimization, Black-Box Optimization

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: We propose a new gradient matching approach to offline model-based optimization with theoretical guarantees

Abstract: Offline design optimization problem arises in numerous science and engineering applications including materials engineering, where expensive online experimentation necessitates the use of in silico surrogate functions to predict and maximize the target objective over candidate designs. Although these surrogates can be learned from offline data, their predictions can be potentially inaccurate outside the offline data regime. This challenge raises a fundamental question about the impact of imperfect surrogate model on the performance gap between its optima and the true oracle optima, and to what extent the performance loss can be mitigated. Although prior work developed methods to improve the robustness of surrogate models and their associated optimization processes, a provably quantifiable relationship between an imperfect surrogate and the corresponding performance gap, and whether prior methods directly address it, remain elusive. To shed more light on this important question, we present a novel theoretical formulation to understand offline black-box optimization, by explicitly bounding the optimization quality based on how well the surrogate matches the latent gradient field that underlines the offline data. Inspired by our theoretical analysis, we propose a principled black-box gradient matching algorithm to create effective surrogate models for offline optimization. Experiments on diverse real-world benchmarks demonstrate improved optimization quality using our approach to create surrogates.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 7067