Archive / INF Seminars / INF_2022_10_28_Svetlana_Berdyugina
USI - Email
 
 
Università
della
Svizzera
italiana
INF
 
 
 
  
 main_banner
 

Inversion problems in solar physics and astrophysics

 
 
 

Host: Prof. Marc Langheinrich

 

Friday

28.10

USI Campus EST, room C1.03, Sector D
09:30 - 10:30
  
 

Svetlana Berdyugina
IRSOL Istituto Ricerche Solari Aldo e Cele Daccò / Euler Institute / Faculty of Informatics, USI
Abstract:
Astrophysical studies of the Sun, distant stars and many other objects in the Universe are largely based on remote sensing techniques, such as imaging, photometry, spectroscopy, polarimetry, interferometry, etc. Inferring physical parameters from such measurements require either direct modeling or inversion methods. Both approaches are time consuming and limited by data noise, physical model assumptions and numerical methods. The common procedure for solving an inversion problem is to employ a minimization technique. However, it is known that due to noise in the data the solution with minimum deviations is too unstable and very noisy. On the other hand, there are a lot of stable and smooth solutions which are compatible with the data within a given level of noise. To select a unique appropriate solution, some additional constraints are necessary, and at present inverse problem solvers differ mostly by the applied constraints (regularizations). For example, the Tikhonov method (TM) searches for the solution with the minimum gradient of the parameters. Similarly, the maximum entropy method (MEM) selects the image with the minimum of information. Such kinds of smoothing are obviously artificial and in fact lead to an apparently acceptable, but distorted solution. Perhaps the least distorted solution can be obtained using the principle component analysis (PCA) combined with the Occam razor principle. Such a co-called Occamian approach was first developed to infer temperature and magnetic field maps of distant stars with magnetic spots similar to those on on the Sun from. It was then employed for inferring solar magnetic fields and surface structures on exoplanets. Considering the current high rate of solar and astrophysical data, we are in search for accelerating and optimizing algorithms to carry out inversions of diverse data simultaneously using machine and deep learning techniques. A collaboration within the Faculty of Informatics in this area will be highly beneficial for solving various problems in solar physics and astrophysics.

Biography:
Prof. Berdyugina is well known for her interdisciplinary research in solar physics, astrophysics and astrobiology using a rarely utilized property of the light – polarization. While employing quantum effects in molecules, magnetic fields and radiation, she pioneered and established several innovative research lines to study magnetism of the Sun, distant stars and relativistic stellar remnants, as well as planets outside the Solar system (exoplanets) and signatures of extraterrestrial life and civilizations which might be detected on Earth-like exoplanets.
Among her prestigious appointments are Professorship at the Federal Institute for Technology ETH Zurich, as a winner of the European Young Investigator (EURYI) Award of the European Science Foundation (ESF) with the support of the Swiss National Science Foundation (SNSF), Professorship at the University of Oulu, Academy Investigator of the Academy of Finland, and NASA Astrobiology Institute Research Fellow at the University of Hawai’i. She is also a winner of a highly competitive European Research Council (ERC) Advance Grant. Along with her new appointments at USI and as IRSOL Director, she holds a full professorship at the University of Freiburg and is the directorship of the Leibniz-Institut für Sonnenphysik (KIS) in Freiburg in Breisgau.