Atinary-technologies/chem-MFBO ·
In this repository, we run an investigation of Multi-fidelity Bayesian Optimization (MFBO) methods and their application to chemistry and materials problems. We first conduct some benchmarks with 2 different acquisition functions and 2 synthetic functions. We also study the parameters favoring the MF approach over the standard single fidelity (SFBO). We run 3 benchmarks in the chemical domain (optimization of Covalent Organic Frameworks (COFs), polarizable molecules and molecule solvation energy). Our study shows how MFBO can effectively reduce the overall cost of optimization tasks and provide guidelines for using this method. 📄 The corresponding paper can be found here Create a new environment Activate the environment and install the package If you want to use the hydra SLURM plugin to launch jobs on a cluster If you want to launch the synthetic functions benchmark, use this command Additionally, the sweep over synthetic functions can be run via The benchmarks for chemistry and mate