Barycentric Kernel for Bayesian Optimization of Chemical Mixture

Chemical-reaction optimization not only increases the yield of chemical processes but also reduces impurities and improves performance resulting products, contributing to important innovations in various industries. This paper presents a novel barycentric kernel for chemical-reaction using Bayesian (BO), powerful machine-learning method designed optimize costly black-box functions. The is specifically tailored as positive definite Gaussian-process surrogate models BO, ensuring stability logarithmic differential operations while effectively mapping concentration space solving problems. We conducted comprehensive experiments comparing proposed with other widely used kernels, such radial basis function (RBF) kernel, across six benchmark functions three Hartmann Euclidean space. results demonstrated kernel’s stable convergence superior these scenarios. Furthermore, highlights importance accurately parameterizing concentrations prevent BO from searching infeasible solutions. Initially reactions, versatile shows promising potential wide range problems, including those requiring meaningful distance metric between mixtures.