Machine learning in electrocatalysis–Living up to the hype?

Abstract

The introduction of machine learning (ML) models in materials science is seen as a paradigm shift in the field. These models enable the thorough exploration of vast material spaces previously deemed beyond the reach of computational studies, thereby accelerating the materials discovery process. In theoretical electrocatalysis, ML models are primarily used as surrogates for, or to complement, more costly ab initio simulations to predict material properties. Herein, the effects ML has had on the field of electrocatalysis are critically reviewed, with particular focus on the degree to which actual progress has resulted from its application. Although the effectiveness of ML in exploring vast material classes is undeniable, the irrational belief in its potential has led to its excessive utilization within the field.