Classification of Fermi-LAT unassociated sources with machine learning in the presence of dataset shifts

Dmitry V. Malyshev

doi:10.1051/epjconf/202531906009

Home/Archive/arXiv

arXiv · 2412.04675

Classification of Fermi-LAT unassociated sources with machine learning in the presence of dataset shifts

Dmitry V. Malyshev

View original source ↗Read source document ↗Export metadata ↓

Abstract

About one third of Fermi Large Area Telescope (LAT) sources are unassociated. We perform multi-class classification of Fermi-LAT sources using machine learning with the goal of probabilistic classification of the unassociated sources. A particular attention is paid to the fact that the distributions of associated and unassociated sources are different as functions of source parameters. In this work, we address this problem in the framework of dataset shifts in machine learning.

Explore related subjects

astro-ph.HE astro-ph.IM

Keep this discovery

BibTeX RIS

Dmitry V. Malyshev. 2024-12-06. Classification of Fermi-LAT unassociated sources with machine learning in the presence of dataset shifts. https://doi.org/10.1051/epjconf%2F202531906009

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

IceCube neutrino point-source searches in the direction of the KM3NeT ultra-high-energy event

While still under construction, the KM3NeT Astroparticle Research with Cosmics in the Abyss (ARCA) detector recorded a $\sim$200 PeV neutrino on February 13th, 2023. This event is the highest-energy neutrino reported. IceCube, a cubic kilometer neutrino detector located at the geographic South Pole, has previously detected neutrinos up to approximately 10 PeV. We search for high-energy neutrinos from the location of the KM3NeT event using 15 years of IceCube data and considering three temporal hypotheses: steady or flaring in time coincidence, or at an arbitrary time. We find no evidence for neutrino emission for any of the studies performed. Correspondingly, we set upper limits on the neutrino flux from a point source in the direction of KM3-230213A. We compare these limits to KM3NeT's estimated flux and show that an astrophysical explanation of this event is strongly constrained for a variety of spectral assumptions for a steady or transient point source with the flux inferred from the single KM3NeT ultra-high-energy event assuming a spectral index of 2.0.

astro-ph.HE↗

arXiv2026

Evidence for the binary nature of the long-period radio transient ASKAP/DART J1832-0911

Long-period transients are a class of periodic pulsed radio source repeating on the minute to hour timescale. Recently, an increasing number of them are being identified as binary systems, specifically white dwarfs with low-mass main-sequence companions. In this work we analyse the most luminous long-period transient discovered to date, ASKAP/DART J1832-0911, with two years of radio data, and propose that it, too, may be a white dwarf system, although in a far more compact orbit than the aforementioned. The pulses are composed of quasi-periodic components which evolve in a systematic way over days and months. The source is highly linearly or elliptically polarised and its brightness enabled very high signal-to-noise measurement of the time-resolved Faraday rotation measure, which was found to vary across pulse phase. The linear polarisation position angle, circular polarised fraction, and spectral index also varied systematically in ways not typical of pulsars and magnetars. We show that an ultra-compact asynchronous polar explains much of the phenomenology of ASKAP/DART J1832-0911, in particular the evolution of the pulse morphology, rotation measure variation, and periodic X-ray emission, although we cannot conclusively prove a binary nature. However, our model makes testable predictions.

astro-ph.HE↗

arXiv2026

Nonbirefringent model of orthogonal polarization modes in radio pulsars - New view on S swing and mode structure in pulsar beam

Two orthogonal polarization modes observed in radio pulsar signals have long been attributed to proper modes of wave oscillation in strongly magnetized plasma. Yet it has been shown recently that they show up readily for extended emission regions that produce incoherently-superposed polarization signal. In this paper we present a two-dimensional polarization model based on incoherent superposition of radio signals. The model involves a single proper mode, say the O mode, but leads to the appearance of two orthogonal polarization tracks and naturally produces the triple form of polarization mode segregation in averaged profiles (central mode flanked on boths sides by another mode), as well as the displacement of modes in latitude, previously inferred from beam mapping. In the case of conal emission regions, the modelled polarization tends to mimic general polarization properties of the rotating vector model (RVM). However, the reason for this is the symmetry of the emission region - not the usual projection of dipolar magnetic azimuths. Thus the emerging RVM parameters reveal geometry of the emission region, not of the dipolar magnetic field. The results strongly support the vital role of nonbirefringent modal effects in radio pulsar profiles. Two proper modes may not be needed to explain observations of two orthogonal polarization tracks.

astro-ph.HE↗