On Deriving Probabilistic Models for Adsorption Energy on Transition Metals using Multi-level Ab initio and Experimental Data

TL;DR

This paper demonstrates that multi-task Gaussian Processes can accurately model adsorption energies on transition metals by effectively combining abundant low-fidelity computational data with scarce high-fidelity or experimental data, improving predictive performance.

Contribution

The paper introduces the application of multi-task Gaussian Processes to integrate multiple data sources for modeling adsorption energies on transition metals, enhancing accuracy over single-task models.

Findings

MT-GP significantly outperforms single-task models.

Combining datasets improves model fidelity.

Method reduces computational and experimental costs.

Abstract

In this paper, we apply multi-task Gaussian Process (MT-GP) to show that the adsorption energy of small adsorbates on transition metal surfaces can be modeled to a high level of fidelity using data from multiple sources, taking advantage of the relatively abundant ''low fidelity" data (such as from density functional theory computations) and small amounts of ''high fidelity" computational (e.g. using the random phase approximation) or experimental data. To fully explore the performance of MT-GP, we perform two case studies - one using purely computational datasets and the other using a combination of experimental and computational datasets. In both cases, the performance of MT-GPs is significantly better than single-task models built on a single data source. This method can be used to learn improved models from fused datasets, and thereby build accurate models under tight computational and experimental budget.