Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MA: Fachverband Magnetismus
MA 19: Characterization and Instrumentation
MA 19.6: Vortrag
Montag, 16. März 2020, 18:30–18:45, HSZ 101
Looking deeper into matter and magnets - scalable recognition of inhomogeneities in video data — Illia Horenko1, •Davi Rodrigues2, Terence O’Kane3, and Karin Everschor-Sitte2 — 1Universita della Svizzera Italiana, Faculty of Informatics, Via G. Buffi 13, TI-6900 Lugano, Switzerland — 2Johannes-Gutenberg University of Mainz, Faculty of Physics, Staudinger Weg 9, 55128 Mainz, Germany — 3Climate Forecasting, CSIRO Oceans and Atmosphere, Castray Esplanade, 7001 Hobart, Tasmania
We present two physics motivated tools which enhances significantly the data extraction from time-discretized measurement data, i.e. video data, compared to state-of-the-art computational methods. We show that these measures detect very subtle material inhomogeneities from magnetic imaging measurements - down to 1% difference in material parameters. We demonstrate the working principle of these measures based on the 2d inhomogeneous Ising model, micromagnetic simulation data as well as experimental magnetization dynamics imaging data.
More generally, we show that these measures - the latent temperature and the latent entropy - reveal information about the system’s memory and its stochasticity, respectively, and they are applicable on a broad range of fields including biology and climate research. Furthermore we prove that they outperform common statistical and machine learning instruments as the iteration costs (scaling and memory requirements) of the algorithm to compute these measures are independent of the data statistics size.