BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:1
UID:UW-Physics-Event-6704
DTSTART:20211020T160000Z
DTEND:20211020T171500Z
DTSTAMP:20260414T192233Z
LAST-MODIFIED:20211012T112631Z
LOCATION:Chamberlin 5280 (Zoom link also available for online particip
 ants who signed up on our mailing list)
SUMMARY:Machine Learning Application for the Event Horizon Telescope\,
  Physics ∩ ML Seminar\, Joshua Yao-Yu Lin \, University of Illinois 
 at Urbana-Champaign
DESCRIPTION:The Event Horizon Telescope (EHT) recently released the fi
 rst horizon-scale images of the black hole in M87. Combined with other
  astronomical data\, these images constrain the mass and spin of the h
 ole as well as the accretion rate and magnetic flux trapped on the hol
 e. An important question for EHT is how well key parameters such as sp
 in and trapped magnetic flux can be extracted from present and future 
 EHT data alone. In the first half of the talk\, we explore parameter e
 xtraction using a neural network trained on high-resolution synthetic 
 images drawn from state-of-the-art simulations. We find that the neura
 l network is able to recover spin and flux with high accuracy. We are 
 particularly interested in interpreting the neural network output and 
 understanding which features are used to identify\, e.g.\, black hole 
 spin. Using feature maps\, we find that the network keys on low surfac
 e brightness feature in particular. In the second half of the talk\, I
  will also mention an ongoing project VLBInet\, in which we propose a 
 data-driven approach to analyze complex visibilities and closure quant
 ities for radio interferometric data with neural networks. Using mock 
 interferometric data\, we show that our neural networks are able to in
 fer the accretion state as either high magnetic flux (MAD) or low magn
 etic flux (SANE)\, suggesting that it is possible to perform parameter
  extraction directly in the visibility domain without image reconstruc
 tion.
URL:https://www.physics.wisc.edu/events/?id=6704
END:VEVENT
END:VCALENDAR