[2Aa01-14] データ駆動表面科学研究部会「データ駆動アプローチ研究の最近の動向」

[2Aa13S] Parameter evaluation of autoencoder for analyzing TOF-SIMS data of three polymers

*Masaru Ito¹, Kazuhiro Matsuda^1,2, Satoka Aoyagi¹ (1. Seikei University, 2. Toray Research Center Inc.)

Time-of-flight secondary ion mass spectrometry (TOF-SIMS) is a strong surface analysis method that can provide 3D images of molecules and chemical structure information. For the analysis of TOF-SIMS data consisting of images and spectra, multivariate analysis methods have been used because TOF-SIMS data interpretation is difficult due to overlapping peaks derived from various molecules. In order to interpret data effectively, it is important to incorporate multiple data analysis methods. In late years, TOF-SIMS data analysis using autoencoder one of the unsupervised learning methods based on artificial neural networks (ANN) has also been reported. However, an optimal parameter setting method for autoencoder has not been established. In this study, we aim to show the basic model of autoencoder for the analysis of TOF-SIMS data by analyzing the TOF-SIMS data of a three-polymer sample using autoencoder.