Visualizing and Interpreting Unsupervised Solar Wind Classifications

• DOI: 10.3389/fspas.2020.553207
• Publication Date: 2020-09-25T16:02:42Z
• Authors: Jorge Amaya; Romain Dupuis; Maria Elena Innoc...; Giovanni Lapenta
• Keywords: Astronomy; Geophysics. Cosmic physics; FOS: Physical sciences; QB1-991; Astronomy & Astrophysics; 7. Clean energy; 01 natural sciences; Physics - Space Physics; 0103 physical sciences; unsupervised; SPACE; CYCLE; ACE; Solar and Stellar Astrophysics (astro-ph.SR); Self-Organizing Maps; 0105 earth and related environmental sciences; autoencoder; COROTATING INTERACTION REGIONS; PCA; Science & Technology; QC801-809; ENERGETIC PARTICLES; Space Physics (physics.space-ph); CORONAL MASS EJECTIONS; machine learning; SLOW; solar wind; Astrophysics - Solar and Stellar Astrophysics; 13. Climate action; Physical Sciences; SHOCK; clustering

One of the goals machine learning is to eliminate tedious and arduous repetitive work. The manual semi-automatic classification millions hours solar wind data from multiple missions can be replaced by automatic algorithms that discover, in mountains multi-dimensional data, real differences properties. In this paper we present how unsupervised clustering techniques used segregate different types wind. We propose use advanced reduction methods pre-process introduce Self-Organizing Maps visualize interpret 14 years ACE data. Finally, show these potentially uncover hidden information, they compare with previous categorizations.