BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.adass.org//adass2026//talk//X93GY8
BEGIN:VTIMEZONE
TZID:AWST
BEGIN:STANDARD
DTSTART:20000101T000000
RRULE:FREQ=YEARLY;BYMONTH=1;UNTIL=20051231T160000Z
TZNAME:AWST
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
END:STANDARD
BEGIN:STANDARD
DTSTART:20070325T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:AWST
TZOFFSETFROM:+0900
TZOFFSETTO:+0800
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20071028T030000
RRULE:FREQ=YEARLY;BYDAY=4SU;BYMONTH=10;UNTIL=20081025T190000Z
TZNAME:AWDT
TZOFFSETFROM:+0800
TZOFFSETTO:+0900
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-adass2026-X93GY8@pretalx.adass.org
DTSTART;TZID=AWST:20261103T141500
DTEND;TZID=AWST:20261103T143000
DESCRIPTION:Radio interferometric imaging is an ill-posed inverse problem b
 ecause the sky brightness distribution is sampled sparsely and irregularly
  in the Fourier (uv) domain. Incomplete uv-coverage contributes to imaging
  artefacts\, uncertain flux recovery\, and distortions of source morpholog
 y\, particularly for faint and extended emission. Many machine-learning ap
 proaches operate on dirty or reconstructed images\, after instrumental eff
 ects and information loss have already been incorporated into the image re
 presentation. Learning directly from complex visibilities offers an altern
 ative: it allows the model to represent the measurement domain continuousl
 y and to infer unsampled Fourier components under a learned prior before c
 onventional image formation.\nWe present the adaptation and high-performan
 ce-computing evaluation of a transformer-conditioned neural field for cont
 inuous visibility-domain modeling. Sparse complex visibilities\, together 
 with their uv coordinates\, are encoded by a Transformer into a latent rep
 resentation. This representation conditions a multilayer perceptron throug
 h feature-wise linear modulation. The resulting neural field can be querie
 d at arbitrary Fourier coordinates and therefore does not restrict the rec
 onstructed visibility function to a fixed grid.\nWe evaluate the approach 
 using simulated observations designed to reproduce key characteristics of 
 LOFAR HBA data. The network is implemented and tested on a multi-GPU HPC s
 ystem. Our study examines model fidelity and generalization to unsampled c
 oordinates and the computational limits imposed by the number of input vis
 ibilities and target image resolution. We also investigate strategies for 
 compressing large visibility datasets into tractable latent representation
 s while retaining information relevant across multiple spatial scales.\nWe
  will present preliminary visibility-modeling results\, computational and 
 memory-scaling benchmarks\, and the methodological developments required t
 o extend the approach to real interferometric observations. These results 
 identify both the potential of continuous visibility representations and t
 he remaining challenges associated with data volume and realistic instrume
 ntal sampling.
DTSTAMP:20261001T111432Z
LOCATION:Banquet Hall
SUMMARY:Toward Scalable Neural Fields for Continuous Visibility Modeling in
  Radio Interferometry - Nicoletta Sanvitale
URL:https://pretalx.adass.org/adass2026/talk/X93GY8/
END:VEVENT
END:VCALENDAR