Of Tokyo Metropolitan University of space physics laboratory Ichinohe Yuto、Yamada Shinya、Naoto Miyazaki(卒業生)When、In research group consisting of four members of Rikkyo University Saito Shinya、For the first time proposed that the introduction of machine learning in the pre-processing of the spectral fit of the X-ray、It is applied to real data、It succeeded in estimating the parameters of the thermal plasma in a single temperature a few percent accuracy。
This、It will greatly expands the possibility to quickly search in the initial value of the fit no bias。The research group、By using the response function of the mirror and micro-calorimeter of the "pupil" satellite、It gave a network of training by using the 10,000 generated data of the spectrum of a single temperature plasma。Applying this network actually in the spectrum of the resulting Perseus cluster of galaxies in the "pupil" satellite、It has demonstrated to be able to estimate the parameters in a few percent of accuracy。
Which it was proposed to use the machine learning as a pre-treatment of the conventional spectral fit、It is expected to be applied to various celestial bodies of the spectrum and science in the future。
- 原論文 (Neural network-based preprocessing to estimate the parameters of the X-ray emission of a single-temperature thermal plasma,en)