第十八期：Applications of Machine Learning in Solar Physics

讲座信息

讲座题目：Applications of Machine Learning in Solar Physics

讲座嘉宾：Dr. Egor ILLARIONOV （Moscow State University）

时间：9月20日上午10点

地点：北航沙河主楼D座10层，D1026会议室

腾讯会议 ：243-386-406，会议密码：301026

讲座摘要

Prominent applications of machine learning raised interest in such models in solar physics. We consider two examples. The first one is a model for coronal holes segmentation in SDO/AIA images. The model was trained to reproduce the segmentation performed by experts (https://observethesun.com/) and is applicable to both solar disk images and synoptic maps. An updated catalog is presented at https://sun.njit.edu. The second example is a model for parametrization of sunspot groups. The idea is to find a compact yet complete representation of the complex geometrical structure of solar groups. We demonstrate that an autoencoder neural network model can be used to find an appropriate representation, and the resulting parameters contain a physical interpretation. Both models and more research projects are available at https://github.com/observethesun.

嘉宾介绍

Dr. Egor Illarionov graduated from Moscow State University in 2013 in the Department of Probability Theory. He received a Ph.D. degree in Solar Physics in 2017. Here is the list of his publications in Google Scholar https://scholar.google.ch/citations?hl=en&user=mJPmiVIAAAAJ . Now he holds lectures and seminars on machine learning, probability theory, and statistics at Moscow State University. He was a visiting scientist at the New Jersey Institute of Technology (USA), NASA Ames Center, Leibniz Institute for Astrophysics in Potsdam (Germany), and Peking University. His research interests include machine learning, solar physics, and MHD.