Abstract
Data analysis is crucial for the interpretation of complex data by sophisticated surface analysis techniques such as time-of-flight secondary ion mass spectrometry (ToF-SIMS). Multivariate analysis is still powerful for such purposes, although recent machine learning techniques may provide more flexible applications. I’d like to talk about machine learning contributions to the ToF-SIMS data analysis and its future development after I briefly introduce how multivariate analysis support the complex data interpretation. Data analysis learning methods are generally divided into three categories, unsupervised learning, supervised learning, and reinforcement learning. For the analysis of surface analysis data, unsupervised learning is mainly useful for extracting features including those related to unknown materials or unknown factors, while supervised learning is helpful for determination, identification and investigation of the relationship between the results by multiple methods. In terms of reinforcement learning, I’d like to discuss for what purposes reinforcement learning is more helpful than other learning methods. In addition, machine learning applications to other surface analysis techniques such as operando hydrogen microscopy based on electron stimulated desorption (ESD), scanning electron microscopy (SEM) and the analysis of the raw data, including all Kikuchi patterns, of electron backscatter diffraction (EBSD) will also be introduced.
