Perspectives on Artificial Intelligence and Machine Learning in Materials Science
| Reference No. | 20210014 |
|---|---|
| Type/Category | Grant for General Research-Workshop（Ⅰ） |
| Title of Research Project | Perspectives on Artificial Intelligence and Machine Learning in Materials Science |
| Principal Investigator | Daniel PACKWOOD （Institute for Integrated Cell-Material Sciences, Kyoto University・Junior Associate Professor） |
| Research Period |
February 4,2022. ～
February 6,2022. |
| Keyword(s) of Research Fields | Materials science, Artificial Intelligence, machine learning, non?linear dynamics, persistent homology, evolutionary algorithms, global optimization algorithms, accelerated discovery of metal alloys, soft materials, complex oxides, density functional theory, Bayesian methods, dynamical systems [PDEs and ODEs], stochastic methods [Monte Carlo] |
| Abstract for Research Report |
Artificial Intelligence has led to a paradigm shift in the investigation of materials science, with machine learning allowing for systematic, informatics-based calculations and predictions and discovery based on material databases, pushing beyond the intrinsic limitations of first-principles calculations. However, successful application requires development of novel methodologies inspired by the frontends of materials development in close synergy between physical science and information technology. This conference will gather an international group of scientists bringing their own distinct perspectives on problems at the intersection of Materials Physics and Information Technology, two areas where interdisciplinary collaborations both at the academic and industrial level are crucial and yet currently in their early phases. Our focus is on analytical (non-linear analysis, ordinary and partial differential equations, optimization, topological data analysis) and computational (Monte Carlo simulation, density functional theory, molecular dynamics) methods and their role in fusing informatics and materials science fields. In order to fuse these fields, powerful optimization methods are required to model the interplay between quantum phenomena that dominate the sub-nano scale and the mesoscale physics ruled by the laws of mechanics, thermodynamics and electromagnetism. The highly non-linear physical laws expressed in terms of systems of non-linear ODEs and PDEs urge the development of new learning methods. Moreover, the development of topological and geometrical methods is necessary to capture and represent underlying structures and statistical laws in large chemical datasets. |
| Organizing Committee Members (Workshop) Participants (Short-term Joint Usage) |
Aleksandar STAYKOV（I2CNER ・Associate Professor） Petros SOFRONIS（I2CNER・Professor, Director） Yasuhide FUKUMOTO（IMI, Kyushu University・Professor） Shigenori FUJIKAWA（I2CNER・Associate Professor） Pierluigi CESANA（IMI, Kyushu University・Associate Professor） Daniel PACKWOOD（Institute for Integrated Cell-Material Sciences, Kyoto University・Junior Associate Professor） |