The University of Tokyo: AI interprets spectra at about 20,000 times faster than experts
~ Without knowledge and craftsmanship, revealing the nature of matter ~
point
- The technology of spectroscopy to investigate the atomic arrangement of materials etc. has been improved, enormous data can be obtained at the same time. Therefore, we can not catch up with manual interpretation by experts.
- Utilizing two kinds of machine learning methods (hierarchical clustering and decision trees), artificial intelligence developed a new method to interpret data. As a result, it can be interpreted at a speed of about 20,000 times compared with the conventional one.
- Without expert knowledge and craftsmanship, you can analyze substances with various spectroscopy at high speed and with high accuracy. For example, it can widely contribute to semiconductor design, battery development, catalyst analysis.
The research results were published on September 6, 2018 (10:00 am UK daylight saving time) online “Scientific Reports” published by UK Nature Publishing Group.
https://www.jst.go.jp/pr/announce/20180906/index.html
Interpretation of material spectra can be data-driven using machine learning – Institute of Industrial Science, the University of Tokyo