We, 14.10 at 3 PM, Maa room; webinar
"I will present clustering techniques applied for two different types of astrophysical data. First of them are datasets about points in an observed region of space, these are the observational galaxy datasets, and catalogues of complex structure elements detected from the spatial distribution of galaxies. The second type of datasets are of astrophysical simulation images. These are images consisting of millions of pixels, each of which represented in an n−dimensional physical feature space. The data analysis problems for these seemingly different two datasets are after all similar: we want to detect clustering of data points. For spatial objects, we were interested in finding spatial clustering or correlation signals, and for the plasma images, we were interested in finding clusters of similar pixels."
This year, Bussov will be defending her doctoral thesis on the same topic.
Join the webinar
Th, 8.10 at 2.15, webinar
Due to rapidly increasing amount of satellite and ground based solar observations, there is urgent demand for fast and scalable data processing methods. Machine learning and especially deep learning is very promising way of solving many tasks in different domains. One such task is for instance solar image restoration, where the aim is to remove from the images aberrations caused by Earth's turbulent atmosphere. There is a potential that in the future such restorations can be done in real time at the site of instrument, so that there is no need to store the raw data. In my talk I will be discussing this and some other machine learning applications in solar physics which can both increase the quality of the clean data as well as computationally outperform methods commonly used in this field.
We, 7.10 at 3 PM, Maa room; webinar
Punyakoti Ganeshaiah Veena
"All the celestial objects, from small asteroids and planets to large galaxies and even the long filaments of the cosmic web are rotating. The origin of rotation is yet to be fully understood, especially in the context of galaxies and their dark matter haloes. According to the classical tidal torque theory, in the early Universe, as matter began to clump together, the resulting anisotropic distribution of matter torqued up proto-galaxies. Simultaneously, the matter overdensities collapsed to form the large-scale filaments, clusters, walls and voids of the cosmic web that we see today. As a result, we expect a correlation between galaxy spin and the cosmic web.
During the talk, Veena will discuss the role of the cosmic web environment in establishing the rotation of haloes and galaxies using large cosmological simulations. She will show that haloes in filaments are in general spinning faster than in other web components. She will also present the correlations between the spin-axis of haloes/galaxies with the orientation of the cosmic filaments that they are growing in and show how these trends evolve with cosmic time and filament properties."
In October, Veena will be defending her doctoral thesis on the same topic. The thesis can be read at the University of Groningen website.
Exhibition of Hubble's space photos
Tartu Observatory, Tõravere