IMM AI lecture by dr. Sergei V. Kalinin (University of Tennessee, Knoxville USA): 'When Machine Learning Meets Physics: Automated Experiment in Microscopy' (Lecture)

Machine learning and artificial intelligence (ML/AI) are rapidly becoming an indispensable part of physics research, with domain applications ranging from theory and materials prediction to high-throughput data analysis. Over the last several years, increasing attention is attracted to the use of AI interacting with physical system as a part of active learning – including materials discovery and optimization, chemical synthesis, and physical measurements. However, proliferation of ML methods in science brings to the forefront the question of how the correlative data-based nature of these methods can be reconciled with active, causal, and hypothesis driven nature of physical sciences, and how ML methods can be used for facilitation of the classical problems such as theory-experiment matching.
Microscopy arguably represents an ideal playground for exploring these questions, since it combines aspects of materials discovery via observation and spectroscopy, physical learning with relatively shallow priors and small number of exogenous variables, and synthesis via controlled interventions. In this presentation, I will discuss recent progress in automated experiment in scanning probe and electron microscopy, ranging from feature to physics discovery via active learning. Here, the strategies based on simple Gaussian Processes often tend to produce sub-optimal results due to the lack of prior knowledge and very simplified (via learned kernel function) representation of spatial complexity of the system. Comparatively, deep kernel learning (DKL) and structured Gaussian Processes methods allow to realize both the exploration of complex systems towards the discovery of structure-property relationship, and enable automated experiment targeting physics (rather than simple spatial feature) discovery. The latter is illustrated via experimental discovery of ferroelectric domain dynamics in piezoresponse force microscopy. For probing physical mechanisms of tip-induced modifications, I will demonstrate the combination of the structured Gaussian process and reinforcement learning, the approach we refer to as hypothesis learning. Here, this approach is used to learn the domain growth laws on a fully autonomous microscope.

The future potential of Bayesian active learning for autonomous microscopes is discussed. These concepts and methods can be extended from microscopy to other areas of automated experiment – including computational discovery and chemical synthesis.

The presentation is followed by discussion and lunch.