Date: Wednesday 27-Sep-2023
Time: 11:30-12:30 pm
Location: PSC 1136
Speaker: Ari Brill (NASA GSFC)
Title: Investigating the Variability of Flaring Gamma-ray Blazars with Stochastic Processes and Deep Learning
Blazars exhibit stochastic flux variability at virtually all observed wavelengths and timescales, from minutes to years, but the physical origin of this variability is poorly understood. For 15 years, the Fermi Large Area Telescope (Fermi-LAT) has conducted regular space-based monitoring of thousands of blazars in the high-energy gamma-ray band. Measurements with Fermi-LAT of several bright flat spectrum radio quasars (FSRQs) have revealed extremely heavy-tailed flux distributions described well by the inverse gamma distribution, as well as complex short-timescale (days or less) variability patterns involving multiple bursts. We explain these observations using a model in which gamma rays are emitted in discrete bursts that are individually unresolved and averaged over within time bins. The statistical model replicates the typical gamma-ray variability properties of FSRQs and BL Lac objects. Often, however, the light curves of flaring blazars exhibit a range of variability patterns that high-level summary statistics or parameterizations may not be able to fully describe. Measurement errors, upper limits, and missing data further complicate the analysis. To interpret complex light curves more effectively, I demonstrate a novel method based on self-supervised deep learning. In this approach, a deep neural network is trained without external supervision to encode a model-independent representation of the stochastic variability structure in the data. The output of the self-supervised model can then be analyzed to extract scientifically relevant information.
Guidelines and format for the CTC Lunch seminars.
Past CTC Lunch seminars.
For other questions, contact the CTC Lunch host: Matt Nixon
SPECIAL ACCOMMODATIONS:
Special accommodations for individuals with disabilities can be made by calling (301) 405-3001. It would be appreciated if we are notified at least one week in advance.
