Generative Pretraining for Black-Box Optimization

Satvik Mehul Mashkaria, Siddarth Krishnamoorthy, Aditya Grover
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:24173-24197, 2023.

Abstract

Many problems in science and engineering involve optimizing an expensive black-box function over a high-dimensional space. In the offline model-based optimization (MBO) setting, we assume access to a fixed, offline dataset for pretraining and a small budget for online function evaluations. Prior approaches seek to utilize the offline data to approximate the function or its inverse but are not sufficiently accurate far from the data distribution. We propose BONET, a generative framework for pretraining a novel model-based optimizer using offline datasets. In BONET, we train an autoregressive model on fixed-length trajectories derived from an offline dataset. We design a sampling strategy to synthesize trajectories from offline data using a simple heuristic of rolling out monotonic transitions from low-fidelity to high-fidelity samples. Empirically, we instantiate BONET using a causally masked Transformer (Radford et al., 2019) and evaluate it on Design-Bench (Trabucco et al., 2022), where we rank the best on average, outperforming state-of-the-art baselines.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-mashkaria23a, title = {Generative Pretraining for Black-Box Optimization}, author = {Mashkaria, Satvik Mehul and Krishnamoorthy, Siddarth and Grover, Aditya}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {24173--24197}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/mashkaria23a/mashkaria23a.pdf}, url = {https://proceedings.mlr.press/v202/mashkaria23a.html}, abstract = {Many problems in science and engineering involve optimizing an expensive black-box function over a high-dimensional space. In the offline model-based optimization (MBO) setting, we assume access to a fixed, offline dataset for pretraining and a small budget for online function evaluations. Prior approaches seek to utilize the offline data to approximate the function or its inverse but are not sufficiently accurate far from the data distribution. We propose BONET, a generative framework for pretraining a novel model-based optimizer using offline datasets. In BONET, we train an autoregressive model on fixed-length trajectories derived from an offline dataset. We design a sampling strategy to synthesize trajectories from offline data using a simple heuristic of rolling out monotonic transitions from low-fidelity to high-fidelity samples. Empirically, we instantiate BONET using a causally masked Transformer (Radford et al., 2019) and evaluate it on Design-Bench (Trabucco et al., 2022), where we rank the best on average, outperforming state-of-the-art baselines.} }
Endnote
%0 Conference Paper %T Generative Pretraining for Black-Box Optimization %A Satvik Mehul Mashkaria %A Siddarth Krishnamoorthy %A Aditya Grover %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-mashkaria23a %I PMLR %P 24173--24197 %U https://proceedings.mlr.press/v202/mashkaria23a.html %V 202 %X Many problems in science and engineering involve optimizing an expensive black-box function over a high-dimensional space. In the offline model-based optimization (MBO) setting, we assume access to a fixed, offline dataset for pretraining and a small budget for online function evaluations. Prior approaches seek to utilize the offline data to approximate the function or its inverse but are not sufficiently accurate far from the data distribution. We propose BONET, a generative framework for pretraining a novel model-based optimizer using offline datasets. In BONET, we train an autoregressive model on fixed-length trajectories derived from an offline dataset. We design a sampling strategy to synthesize trajectories from offline data using a simple heuristic of rolling out monotonic transitions from low-fidelity to high-fidelity samples. Empirically, we instantiate BONET using a causally masked Transformer (Radford et al., 2019) and evaluate it on Design-Bench (Trabucco et al., 2022), where we rank the best on average, outperforming state-of-the-art baselines.
APA
Mashkaria, S.M., Krishnamoorthy, S. & Grover, A.. (2023). Generative Pretraining for Black-Box Optimization. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:24173-24197 Available from https://proceedings.mlr.press/v202/mashkaria23a.html.

Related Material