Please consider attending the following: MATERIALS ENGINEERING SEMINAR "Information Acquisition Approaches for Active Learning and its Applications to Materials Science" By Juan Carlos Verduzco Gastelum Purdue MSE Preliminary Exam Advisors: Professor Esteban Marinero and Professor Alejandro Strachan ABSTRACT Development and application of machine learning models for the prediction of properties in materials science have gained momentum in the past decades. These algorithms boast the ability to make accurate, fast, and inexpensive predictions that can guide the design of experiments to the most likely candidates for success. However, predictions are entirely dependent on the amount and quality of the data used to train the fitted model. Datasets, both in experimental setups and in computational simulations for materials science, are small, and the generation of new entries remains a time consuming and expensive task. Active learning aims to inform new experiments to produce a gain in information and has been successful in reducing the number of experiments needed to discover a material with an optimized property. This work describes a new paradigm in the discovery of materials based on data science, analyzes the role of modeling in active learning applications for materials science, and benchmarks different acquisition functions in a Bayesian global optimization framework. Date: Tuesday, December 10, 2019 Time: 10:30 A.M. Place: ARMS 1028 School of Materials Engineering Purdue University Neil Armstrong Hall of Engineering 701 West Stadium Ave. Room 2200 West Lafayette, IN 47907 765-494-4105 [PU150]
