Abstract: The arrival of materials science data infrastructures in the past decade has ushered in the era of data-driven materials science based on artificial intelligence (AI) algorithms, which has facilitated breakthroughs in materials optimization and design. Here, Bayesian optimization is a particularly versatile tool for prediction tasks.

One well-known aspect of materials simulations is a hierarchy of computational approaches available. We combined multi-output Gaussian processes with cost-aware acquisition functions to carry out multi-fidelity Bayesian optimization: these techniques have the potential to considerably accelerate scientific discovery.

Speaker: Milica Todorović is an Assistant Professor of Materials Engineering at the University of Turku. She gained a DPhil in Materials Science at the University of Oxford, then specialised in high-performance computations of materials in Japan and Spain. After arriving to Aalto University, Milica started collaborating with the Finnish Center for Artificial Intelligence (FCAI) and has organised workshops and graduate courses on Machine Learning for Materials Science. Her research focuses on interfacing artificial intelligence algorithms with materials science simulations and experiments, with the aim to optimise material functionality and performance in devices.

Affiliation: University of Turku

Place of Seminar:  Zoom