BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Physics-Informed Machine Learning for precise and accurate weak le
 nsing shear estimation
DTSTART:20260820T190000Z
DTEND:20260820T200000Z
DTSTAMP:20260628T004200Z
UID:indico-event-32371@indico.bnl.gov
CONTACT:xli6@bnl.gov
DESCRIPTION:Speakers: Shurui Lin (UIUC)\n\nWeak gravitational lensing has 
 served as an important probe for large-scale structure and cosmology for d
 ecades. Stage-IV surveys require both sub-percent calibration accuracy and
  high statistical precision for WL shear estimation\, yet traditional esti
 mators struggle with realistic galaxy complexity while machine-learning me
 thods often introduce biases.\nI will introduce a physics-informed machine
 -learning approach that combines a fully D₄-equivariant convolutional ne
 ural network (D₄CNN) with a score-matching technique for optimal shear e
 stimation. The D₄CNN enforces symmetry under rotations and reflections\,
  eliminating even-order shear biases by construction\, while Analytical Ca
 libration (AnaCal) provides precise\, gradient-based self-calibration.\nTo
 gether with modern denoising-score-matching framework\, our method achieve
 s multiplicative biases consistent with zero at the ∼10⁻⁴ level\, we
 ll within the requirement of Stage IV surveys like LSST\, and reduces shap
 e noise by ∼20% relative to the classical baseline\, (equivalent to ∼4
 0% increase in observation time)\, providing a principled and practical ma
 chine-learning pathway toward optimal shear estimation for Stage-IV survey
 s.\nZoom link: https://bnl.zoomgov.com/j/16001410126?pwd=LO1VgLQipgjIfQpO
 lKB6szuFxr7L7g.1\n\nhttps://indico.bnl.gov/event/32371/
LOCATION:Small Seminar Room (Building 510)
URL:https://indico.bnl.gov/event/32371/
END:VEVENT
END:VCALENDAR