イベント
第115回先端計測オープンセミナー
計算・計測データ解析のための機械学習技術の応用研究
安藤康伸 (産総研 機能材料コンピュテーショナルデザイン研究センター 研究員)
2017年12月11日（月）15:00～16:00
会場/Venue：
千現地区 研究本館8階 中セミナー室
Sengen Main Bldg. 8F Middle Seminar Room
Sengen Main Bldg. 8F Middle Seminar Room
講演者/Speaker：
安藤康伸 (産業技術総合研究所 機能材料コンピュテーショナルデザイン研究センター 研究員)
Yasunobu ANDO, Researcher (National institute of Advanced Science and Technology, Research Center for Computational Design of Advanced Functional Materials (CD-FMat))
Yasunobu ANDO, Researcher (National institute of Advanced Science and Technology, Research Center for Computational Design of Advanced Functional Materials (CD-FMat))
表題/Title：
計算・計測データ解析のための機械学習技術の応用研究
Study of machine-learning application for high-productive research activity in material science
Study of machine-learning application for high-productive research activity in material science
講演要旨/Abstract：
材料開発の現場では所望の新機能を有する材料の高速開発技術の取得は避けられない課題となってきている。一方で、そもそも現在使われている高性能材料の機能がどのようなメカニズムで発現しているかも非自明な領域が多く、材料探索及び基礎物性研究双方の立場から、 計測・計算双方から広大な物質データ空間からの情報抽出技術の開発が強く望まれている。本講演では、現在注目を集めている機械学習技術を適用してのデータ駆動による物質・材料に対する適用事例を紹介し、当該分野・情報科学分野の連携課題について議論する。
To accelerate research for finding new functional materials, research activity with computer simulation combined with machine learning technology has attracted interests all over the world. Although resent trend is handled by computational research community mainly, machine learning is also promising to make an impact on experimental research field. From these circumstances, we believe that machine learning will become indispensable tools for our high-productive research activity. In this seminar, I would like to introduce our recent studies how to apply machine learning such as compress sensing, unsupervised learning, and neural network modeling for interatomic potentials, to analyze our results obtained by first-principles simulation and experiments. I would also like to discuss issues to apply the machine learning techniques to our research activity.
To accelerate research for finding new functional materials, research activity with computer simulation combined with machine learning technology has attracted interests all over the world. Although resent trend is handled by computational research community mainly, machine learning is also promising to make an impact on experimental research field. From these circumstances, we believe that machine learning will become indispensable tools for our high-productive research activity. In this seminar, I would like to introduce our recent studies how to apply machine learning such as compress sensing, unsupervised learning, and neural network modeling for interatomic potentials, to analyze our results obtained by first-principles simulation and experiments. I would also like to discuss issues to apply the machine learning techniques to our research activity.