HypBO: Accelerating Black-Box Scientific Experiments Using Experts’ Hypotheses

Authors: Abdoulatif Cissé, Xenophon Evangelopoulos, Sam Carruthers, Vladimir V. Gusev, Andrew I. Cooper

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

Reproducibility Variable Result LLM Response
Research Type Experimental We validate the performance of our method on a range of synthetic functions and demonstrate its practical utility on a real chemical design task where the use of expert hypotheses accelerates the search performance significantly.
Researcher Affiliation Academia 1Department of Chemistry, University of Liverpool, England, UK 2Leverhulme Research Centre for Functional Materials Design, University of Liverpool, England, UK 3Department of Computer Science, University of Liverpool, England, UK {abdoulatif.cisse, evangx, sgscarru, vladimir.gusev, aicooper}@liverpool.ac.uk
Pseudocode Yes Algorithm 1 Hypothesis Bayesian Optimization (Hyp BO)
Open Source Code Yes Reproducibility details are available in the SM and the source code can be found at https://github.com/Ablatif6c/Hyp BO.
Open Datasets Yes We fitted this model against a total ground truth dataset of 1119 experimental observations supplied by the authors of [Burger et al., 2020].
Dataset Splits No The paper describes experiments on synthetic functions and a simulated chemical space, but does not provide explicit details on training, validation, or test dataset splits (e.g., percentages or counts) for the BO problems themselves. The GPR model for the chemical space was 'fitted against a total ground truth dataset of 1119 experimental observations' without specifying explicit splits for that fitting process.
Hardware Specification No The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running its experiments.
Software Dependencies No The paper does not provide specific software dependencies, such as library names with version numbers (e.g., 'PyTorch 1.9', 'Python 3.8'), needed to replicate the experiment.
Experiment Setup Yes For all experiments, we use preset hyperparameters for Hyp BO. We set the lower level limit lmax to 2, the upper level limit umax to 5, the number of locally optimal samples T to 1, and the growth rate γ to 0. All experiments are warm-started with five initial points except for the photocatalytic hydrogen production experiment with mixed hypotheses, whose initial sample count is 10.