Machine learning (ML) accelerates materials discovery by mapping structure-property relationships, but its efficiency is often limited by the need for massive datasets. To address this, we framed the search for high-performance nanoporous materials (MOFs/COFs) as a Bayesian Optimization (BO) problem.

This framework provides a data-efficient, computationally accessible route for navigating massive design spaces to find the most promising materials for adsorption and diffusion applications.