Physics Dept Seminar


February 11th, TUESDAY (*SPECIAL DAY*)


How Machine Learning Helps to Analyze Solar Disk Images


Prof. Egor Illarionov

Moscow Univ.

(Solar Physics, Host: Nita)


Time: 11:45 am - 12:45 pm with 11:30 am tea time

Room: ECE 202


Interpretation of solar disk observations is a challenging task. To highlight the problems we will start with a review of several databases of source solar disk images and catalogues of solar features. Then we will discuss how to collect a proper dataset to proceed with machine learning models. The core part of the talk will be a demonstration of recent results in segmentation of coronal holes and sunspots in SDO images with a convolutional neural network of the U-Net type. Finally, we will discuss a necessity of common guidelines in doing of machine learning researches in solar physics and what it could be.


Biography Dr Egor Illarionov is Assistant Professor at the Department of Mathematics and Mechanics of Moscow State University (Russia). He received his PhD in 2017 from Moscow University in the area of application on mathematical methods to solar dynamo modeling and solar image analysis.