BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//pretalx.adass.org//adass2026//talk//LLETHA
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-LLETHA@pretalx.adass.org
DTSTART;TZID=AWST:20261103T155100
DTEND;TZID=AWST:20261103T160300
DESCRIPTION:We present an ongoing investigation of a non-learnable approach
  for pulsar-signal detection using handcrafted 2D line filters. The method
  applies Gaussian ridge filters across multiple scales and orientations to
  identify the characteristic dispersed structures of pulsars in dynamic sp
 ectra. Small random perturbations are introduced into the filters\, creati
 ng a noise-augmented filter bank whose responses are subsequently aggregat
 ed to reinforce consistent signal structures while suppressing uncorrelate
 d fluctuations. \nThe filter responses are processed through a sequence of
  sigmoid activation\, nonlinear transformation\, multi-scale summation\, b
 inarization\, orientation alignment\, and peak-prominence selection. This 
 adaptive processing identifies orientations that provide consistent eviden
 ce for a pulsar signal without relying on learned convolutional kernels.\n
 The current results on the B0531+21 dataset demonstrate the potential of t
 his handcrafted filtering approach\, with an overall F1-score above 90%.
DTSTAMP:20261116T154857Z
LOCATION:Seminar Room 2
SUMMARY:Pulsar Signal Segmentation and Analysis using Custom Convolutional 
 Filter Stacks - Tanumoy Saha
URL:https://pretalx.adass.org/adass2026/talk/LLETHA/
END:VEVENT
END:VCALENDAR
