Bayesian Optimization for Metallophotocatalysis Formulations

This repository comprises scripts used in the research paper titled 'Sequential Closed-Loop Bayesian Optimization as a Guide for Organic Molecular Metallophotocatalyst Formulations Discovery'.

Code Structure

Overview

The Bayesian Optimization instance BayesOptimizer, available in src.BayesOpt.optimizer, is employed for the optimization of organic catalysts and the reaction conditions.

The Gaussian processes instance GaussianProcessRegressor, created in src.BayesOpt.GPR, is used as the surrogate model of the optimization of reaction conditions.

The Jupyter notebooks in workflows are the original record of the optimization of CNPs and reaction conditions. The kernel matrix of designed reaction conditions need to be built before running the optimization.

Major functions

Function	Description
src.data_pretreatment.ks_selection	The Kennard-Stone algorithm used for selecting represent subset.
src.data_pretreatment.cal_fingerprints	Calculating the Fingerprints of given chemical SMILES.
src.BayesOpt.BayesOptimizer.ask	Query one point at which objective should be evaluated.
src.BayesOpt.BayesOptimizer.parallel_ask	Query several points at which objective should be evaluated.
src.BayesOpt.BayesOptimizer.tell	Recording evaluated points of the objective function.
src.BayesOpt.acquisition_function	Computing the acquisition function.
src.BayesOpt.GaussianProcessRegressor.fit	Fit Gaussian process regression model.
src.BayesOpt.GaussianProcessRegressor.predict	Predict using the Gaussian process regression model.

Author

Yu Che