Jagiellonian University

About: Jagiellonian University is a education organization based out in Krakow, Poland. It is known for research contribution in the topics: Population & Catalysis. The organization has 17438 authors who have published 44092 publications receiving 862633 citations. The organization is also known as: Academia Cracoviensis & Akademia Krakowska.

Exploring cosmic origins with CORE: Survey requirements and mission design

Abstract: Future observations of cosmic microwave background (CMB) polarisation have the potential to answer some of the most fundamental questions of modern physics and cosmology, including: what physical process gave birth to the Universe we see today? What are the dark matter and dark energy that seem to constitute 95% of the energy density of the Universe? Do we need extensions to the standard model of particle physics and fundamental interactions? Is the ΛCDM cosmological scenario correct, or are we missing an essential piece of the puzzle? In this paper, we list the requirements for a future CMB polarisation survey addressing these scientific objectives, and discuss the design drivers of the COREmfive space mission proposed to ESA in answer to the "M5" call for a medium-sized mission. The rationale and options, and the methodologies used to assess the mission's performance, are of interest to other future CMB mission design studies. COREmfive has 19 frequency channels, distributed over a broad frequency range, spanning the 60–600 GHz interval, to control astrophysical foreground emission. The angular resolution ranges from 2' to 18', and the aggregate CMB sensitivity is about 2 μK⋅arcmin. The observations are made with a single integrated focal-plane instrument, consisting of an array of 2100 cryogenically-cooled, linearly-polarised detectors at the focus of a 1.2-m aperture cross-Dragone telescope. The mission is designed to minimise all sources of systematic effects, which must be controlled so that no more than 10−4 of the intensity leaks into polarisation maps, and no more than about 1% of E-type polarisation leaks into B-type modes. COREmfive observes the sky from a large Lissajous orbit around the Sun-Earth L2 point on an orbit that offers stable observing conditions and avoids contamination from sidelobe pick-up of stray radiation originating from the Sun, Earth, and Moon. The entire sky is observed repeatedly during four years of continuous scanning, with a combination of three rotations of the spacecraft over different timescales. With about 50% of the sky covered every few days, this scan strategy provides the mitigation of systematic effects and the internal redundancy that are needed to convincingly extract the primordial B-mode signal on large angular scales, and check with adequate sensitivity the consistency of the observations in several independent data subsets. COREmfive is designed as a "near-ultimate" CMB polarisation mission which, for optimal complementarity with ground-based observations, will perform the observations that are known to be essential to CMB polarisation science and cannot be obtained by any other means than a dedicated space mission. It will provide well-characterised, highly-redundant multi-frequency observations of polarisation at all the scales where foreground emission and cosmic variance dominate the final uncertainty for obtaining precision CMB science, as well as 2' angular resolution maps of high-frequency foreground emission in the 300–600 GHz frequency range, essential for complementarity with future ground-based observations with large telescopes that can observe the CMB with the same beamsize.

Statistical properties of random density matrices

Abstract: Statistical properties of ensembles of random density matrices are investigated. We compute traces and von Neumann entropies averaged over ensembles of random density matrices distributed according to the Bures measure. The eigenvalues of the random density matrices are analysed: we derive the eigenvalue distribution for the Bures ensemble which is shown to be broader then the quarter-circle distribution characteristic of the Hilbert–Schmidt ensemble. For measures induced by partial tracing over the environment we compute exactly the two-point eigenvalue correlation function.

Galaxy–galaxy weak lensing in the Sloan Digital Sky Survey: intrinsic alignments and shear calibration errors

Abstract: Galaxy-galaxy lensing has emerged as a powerful probe of the dark matter haloes of galaxies, but is subject to contamination if intrinsically aligned satellites of the lens galaxy are used as part of the source sample. We present a measurement of this intrinsic shear using 200 747 lens galaxies from the Sloan Digital Sky Survey (SDSS) spectroscopic sample and a sample of satellites selected using photometric redshifts. The mean intrinsic shear at transverse separations of 30-446 h -1 kpc is constrained to be -0.0062 < Δγ < +0.0066 (99.9 per cent confidence, including identified systematics), which limits contamination of the galaxy-galaxy lensing signal to at most ∼15 per cent on these scales. We present these limits as a function of transverse separation and lens luminosity. We furthermore investigate shear calibration biases in the SDSS, which can also affect galaxy-galaxy lensing, and conclude that the shear amplitude is calibrated to better than 18 per cent. This includes noise-induced calibration biases in the ellipticity, which are small for the sample considered here, but which can be more important if low signal-to-noise ratio or poorly resolved source galaxies are used.

On Loss Functions for Deep Neural Networks in Classification

Abstract: Deep neural networks are currently among the most commonly used classifiers. Despite easily achieving very good performance, one of the best selling points of these models is their modular design - one can conveniently adapt their architecture to specific needs, change connectivity patterns, attach specialised layers, experiment with a large amount of activation functions, normalisation schemes and many others. While one can find impressively wide spread of various configurations of almost every aspect of the deep nets, one element is, in authors' opinion, underrepresented - while solving classification problems, vast majority of papers and applications simply use log loss. In this paper we try to investigate how particular choices of loss functions affect deep models and their learning dynamics, as well as resulting classifiers robustness to various effects. We perform experiments on classical datasets, as well as provide some additional, theoretical insights into the problem. In particular we show that L1 and L2 losses are, quite surprisingly, justified classification objectives for deep nets, by providing probabilistic interpretation in terms of expected misclassification. We also introduce two losses which are not typically used as deep nets objectives and show that they are viable alternatives to the existing ones.