Colloquia:
3. Dr. Prof. Yong-Jae Moon, (School of Space Exploration, Kyung Hee University), Title: What kinds of problems in solar physics and space weather are well solved by deep learning?, Nov 19, 2024, 4pm MST
Abstract: In this talk, I introduce our recent deep learning applications to solar and space weather data. We have successfully applied novel deep learning methods to the following applications: (1) image translations between solar images, (2) generation of UV/EUV images and magnetograms from Galileo sunspot drawings, (3) denoising solar magnetograms using supervised learning, (4) super resolution of solar magnetograms, (5) near real-time construction of solar coronal parameters based on MAS, (6) improvement of global IRI TEC maps using IGS TEC ones, (7) one-day forecasting of global TEC maps, (8) flare classification by CNN and visual explanation by attribution methods, (9) forecasting GOES solar X-ray profiles, and (10) three-day forecast of solar wind speed using solar images. We present major results and discuss what kinds of problems in solar physics and space weather are well solved by deep learning.
2. Dr. Yang Chen (University of Michigan), Title: Data-Analytic Opportunities and Challenges in Solar Eruption Forecasting, June 11, 2024, 1pm MDT
Abstract: In recent years, there has been a growing awareness of space weather impacts on critical infrastructure in the civilian, commercial, and military sectors. The interest will continue to grow as we gain a better understanding of the physical processes of the Sun and their effects here on Earth and in space itself. Among all space weather events, the solar flare is a relatively intense, localized emission of electromagnetic radiation in the Sun’s atmosphere. Flares occur in active regions and are often, but not always, accompanied by coronal mass ejections, solar particle events, and other eruptive solar phenomena. In recent years, machine learning models have been employed to forecast solar eruptions, including strong solar flare events. In this talk, I will discuss data-analytic opportunities and challenges in solar eruption forecasting, given the availability of massive solar imaging data and other relevant data products. Scientifically, it is essential to (a) identify the solar active regions that have high potential to erupt in an automated fashion, (b) extract features from observed solar images using principled statistical algorithms, (c), most importantly, provide a probabilistic forecast for the eruption time, magnitude and magnetic field configuration, and (d) facilitate new understandings of the mechanism/physics of disruptive space weather events. In particular, I will present one of our novel statistical models, the Tensor Gaussian Process with Contraction, for solar flare forecasting, combining data from various types and sources. I will conclude with ongoing and future work on operational solar eruption forecasting.
Please find relevant papers and presentation slides on my website here: https://yangchenfunstatistics.github.io/yangchen.github.io//
1. Dr. Azim Ahmadzadeh (University of Missouri-St. Louis), Title: Evaluation Challenges in Detection and Classification ML Models, May 21, 2024, 1pm MDT
Abstract: In this talk, I will delve into one of the most persistent challenges within the machine learning (ML) domain: the effective evaluation of models. The ML community has long been aware of these evaluation challenges, yet progress in addressing them has been notably slow. Additionally, there remains a significant gap in awareness among ML users about the biases introduced by superficial evaluation measures and frameworks. I scratch the surface of this topic by highlighting two critical issues, and offer practical solutions to enhance the understanding and application of ML evaluation, particularly in space weather research.