Скачать с ютуб Machine learning in materials science в хорошем качестве

Machine learning in materials science

A lecture given by Harry Bhadeshia at the University of Birmingham, Department of Materials Science and Metallurgy. Long before machine learning and artificial intelligence became throwaway terms and on occasions seen as threats to our very existence, there were useful attempts to take advantage of what are essentially mathematical methods to deal with complex problems in science. In the context of materials science, the first paper seems to have been published in 1991 [1]. Professor Harry Bhadeshia's own interest was stimulated in 1992, when he did not understand a presentation made at a conference, and followed this up on returning to Cambridge, where he discovered David MacKay (a leader in information theory), then in the Physics Department. They started working together and published the first paper in 1995 [2]. The method turned out to be so powerful that they published 14 papers together in unfunded research over a period of 24 years. In his lecture, Professor Sir Harry Bhadeshia will deal explicitly with the following: (a) A simple and transparent explanation of the method. (b) Question whether there is any intelligence involved. (c) Show how the technique has led to remarkable PREDICTIONS that have been subsequently been verified experimentally. (d) Describe the best way of disseminating the outcomes. [1] Ghaboussi, J., J. H. Garrett Jr, and Xiping Wu. "Knowledge-based modeling of material behavior with neural networks." Journal of Engineering Mechanics 117.1 (1991): 132-153. [2] Bhadeshia, H. K. D. H., D. J. C. MacKay, and L-E. Svensson. "Impact toughness of C–Mn steel arc welds–Bayesian neural network analysis." Materials Science and Technology 11.10 (1995): 1046-1051. Harry Bhadeshia is the Tata Steel Professor of Metallurgy at the University of Cambridge, with primary interest in the science that leads to novel alloys of iron.