[17p-P10-1~93] 10 Spintronics and Magnetics（Poster）

[17p-P10-65] Feature extraction of coercivity using topological machine learning

〇Masato Kotsugi^1,3, Takumi Yamada^1,3, Shingo Suzuki¹, Yuta Suzuki¹, Ippei Obayashi², Hiroaki Hiraoka^2,3,4 (1.Tokyo Univ. of Science, 2.AIMR Tohoku Univ., 3.MI2I-NIMS, 4.AIP center RIKEN)

Keywords:Machine learning, magnetic domain, Topological data analysis

We visualized the origin of coercivity in the magnetic domain by using machine learning with topological data analysis. We utilized “Persistent homology”, which is known as a compact descriptor for characterizing multiscale topological features in data, for feature extraction from magnetic domain structure. In the combination of principal component analysis (PCA), we could visualize point clouds contributing to coercivity in magnetic domain structure.