Machine learning and data-driven techniques are quickly being adopted to accelerate materials research in a variety of ways. In this tutorial, we introduce materials scientists to a wide variety of machine learning topics, which have found utility in real-world materials research. We will review fundamental topics in machine learning, including supervised and unsupervised learning, reinforcement learning, and Bayesian techniques and optimization. We will also cover practical tools and techniques for handling experimental data, in addition to extracting the relevant information from such data to make the applications of machine learning methods possible. After the tutorial, participants will have a broad understanding of machine learning in general, as well as concrete example applications of the topics to materials science problems. No previous knowledge of machine learning will be required.
8:30 am -- CANCELLED
Data Fundamentals, Experimental Data and Computation
Filtering, statistical tools for experimental data, feature extraction and engineering.
9:45 am BREAK
Daniel V. Samarov
Regression and classification models and techniques including regularized least squares, support vector machines, neural networks, ensemble learning, gaussian processes.
Aaron Gilad Kusne
Clustering, similarity measures, latent variable analysis, spectral unmixing, matrix factorization.
2:45 pm BREAK
Sequential Experimental Design and Reinforcement Learning
Bayesian optimization and experimental design, belief models, decision policies, Markov decision processes.
- Aaron Gilad Kusne, National Institute of Standards and Technology
- Daniel V. Samarov, National Institute of Standards and Technology
- Alexander Hexemer, Lawrence Berkeley National Laboratory
- Kristofer Reyes, University at Buffalo, The State University of New York