Position: Leverage Foundational Models for Black-Box Optimization

Authors: Xingyou Song, Yingtao Tian, Robert Tjarko Lange, Chansoo Lee, Yujin Tang, Yutian Chen

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this position paper, we frame the field of blackbox optimization around sequence-based foundation models and organize their relationship with previous literature. We discuss the most promising ways foundational language models can revolutionize optimization...
Researcher Affiliation Industry 1Google DeepMind 2Sakana AI.
Pseudocode Yes Problem 1 Experimental Design Problem Require: Problem meta-description m M, search space X, and feedback function f F, Algorithm A Initialize A using all provided information. Initialize history h [] for t=1, 2, . . . until end condition is met do Generate a suggestion: xt A(m, h1:t 1) Receive feedback: yt f(xt) Update history: h concatenate(h, xt, yt). end for Return h1:t as H
Open Source Code No The paper is a position paper discussing concepts and future directions, and does not provide an explicit statement or link for open-source code related to its own methodology.
Open Datasets No The paper discusses the availability of datasets like Google Vizier and Ax, as well as smaller public datasets (Eggensperger et al., 2021; Trabucco et al., 2022) and Open ML (Bischl et al., 2021) in the context of learned BBO algorithms, but it does not report on experiments using these datasets or provide access information for data used in its own work.
Dataset Splits No The paper does not report on new experiments or dataset usage, therefore no training/validation/test dataset splits are provided.
Hardware Specification No The paper does not report on new experiments, and therefore no specific hardware specifications for running experiments are provided.
Software Dependencies No The paper does not report on new experiments, and therefore does not list specific software dependencies with version numbers.
Experiment Setup No The paper does not report on new experiments, and therefore no specific experimental setup details or hyperparameters are provided.