Download all files (84.5MB)

コレクション

引用

説明:

The prediction performance of the automatic threshold estimation of photoelectron yield spectroscopy using machine learning and least squares regression (fitting) was verified using 82 measured data. The correct answer rate was greater than 80％ for machine learning and less than 50％ for fitting, within an error range of ±0.3 eV with respect to the correct answer (the result of human spectrum analysis). To further improve the correct answer rate, it is necessary to change the energy range and energy step of the measured data because increase in the nonlinear intensity at the highenergy side of the spectrum is considered to affect the automatic estimation. The estimation did not change with changes in the energy step of the data. However, when the energy range in the data was changed, the estimation improved. If the threshold is expected to be less than 6.0 eV, the prediction is improved by using an energy range of 4.2–6.2 eV.

権利情報:

キーワード: threshold, machine learning, photoelectron yield spectroscopy

刊行年月日: 2020-06-10

出版者: The Japan Society of Vacuum and Surface Science

掲載誌:

研究助成金:

原稿種別: 査読前原稿 (Author's original)

MDR DOI: https://doi.org/10.48505/nims.1440

公開URL: https://doi.org/10.1380/vss.63.270

関連資料:

その他の識別子:

連絡先:

更新時刻: 2024-01-05 22:12:44 +0900

MDRでの公開時刻: 2021-08-13 01:20:02 +0900