Fig.1 (Left) the original image, and (right) the optimizing result using PCA.[1] Fumihiko Uesugi. Novel image processing method inspired by wavelet transform. Micron, 168, 103442 (2023) Fig. 1 Dimensional reductionof Fermi surface of CMGG83Poster Award NomineeP5-05A feature Mining Method Composed of Wavelet Filtering and PCAFumihiko Uesugi1, Koji Harano2, and Koji Kimoto21 Electron microscopy unit, National Institute for Materials Science (NIMS)2 Center for Basic Research on Materials, National Institute for Materials Science (NIMS)The image processing method developed by Uesugi can perform feature extraction, image enhancement, and noise filtering by appropriately selecting and superimposing elements expanded by a concentric mother wavelet (CMW) [1]. In the feature extraction, size information of an object is extracted as wavelet coefficients using the CMW with the various radius, selecting, and reconstructed multiplying the coefficients and CMW. P5-06Information Extraction from Fermi Surfaces Using Unsupervised Machine LearningDaichi Ishikawa1, Kentaro Fuku1,Yoshio Miura2,3,Yasuhiko Igarashi4, Yuma Iwasaki5, Yuya Sakuraba2,Koichiro Yaji5, Alexandre Lira Foggiatto1, Varadwaj Arpita1, Naoka Nagamura5, and Masato Kotsugi11 Department of Material Science and Technology, Tokyo University of Science2 Research Center for Magnetic and Spintronics Materials, National Institute for Materials Science (NIMS)3 Electrical Engineering and Electronics, Kyoto Institute of Technology4 Institute of System and Information Engineering, University of Tsukuba5 Center for Basic Research on Materials, National Institute for Materials Science (NIMS)Fermi surface is crucial information for the designing various functions in spintronics devices. Particularly, electron states such as Weyl point and nodal line on Fermi surface contribute to spin polarization and anomalous Nernstian effects. In this study, we applied machine learning to the Fermi surface of Heusler alloy Co2MnGaxGe1-x (CMGG) and visualized the regions contributing to physical properties. We could confirm significant “jumps” in certain compositions in the reduced two-dimensional space. These jumps corresponded to compositions with varying spin polarization rates, and to compositions where the Weyl points appeared in the Fermi level. These results demonstrate the success of using unsupervised machine learning to reduce the dimensionality of the complex Fermi surface and visualize it in data space, allowing for the automatic extraction of noteworthy compositions and features.In this study, we tried to optimize the feature extraction by performing principal component analysis (PCA) on the wavelet coefficient data. Extracting the “feature” as loading vectors, the target object in the image can be reconstructed as score vectors. If feature extraction is possible, it is expected that the target object in another image will be selectively enhanced. The bottom right of the figure shows the result of feature extraction by PCA. The polymer image was reconstructed without losing the fine structure (Fig.1).
元のページ ../index.html#83