While tremendous advances have been achieved in data science, the resulting models are empirical in nature. In addition, data science has focused on big data. However, in many applications, some of the data is often small, for which, typical statistics approaches and machine learning do not work. Furthermore, even massive data is often correlated and thus low dimensional. Data-based empirical models can violate conservation laws, e.g., momentum conservation, and constraints, e.g., materials stability limits, both of which are commonplace in engineering applications. It is therefore essential to integrate data with physics-based models and expert knowledge. These challenges combined define a critical need for developing a next generation modeling platform that integrates the aforementioned pillars leading to chemical artificial intelligence (AI). These concepts will be illustrated with examples focusing on electrocatalysis and specifically on the oxygen reduction reaction.

Artificial intelligence for chemical sciences

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.