BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Neural geometric PDEs on neural implicit surfaces [CRC 1720: Early
  Career Seminar]
DTSTART:20251202T131500Z
DTEND:20251202T141500Z
DTSTAMP:20260415T093000Z
UID:indico-event-833@math-events.uni-bonn.de
CONTACT:sfb1720-seminar@uni-bonn.de
DESCRIPTION:Speakers: Samuel Weidemaier\n\nThis seminar talk presents work
  on a neural framework for solving geometric partial differential equation
 s (PDEs) on neural implicit manifolds. The key ingredient is a new method 
 for constructing accurate signed distance functions (SDFs) from unoriented
  point clouds\, replacing the standard Eikonal-based approach with a neura
 l variational heat method. This idea results in a two-step method: first a
 pproximating unsigned gradient directions\, and then recovering the correc
 t normal orientation. Both steps are shown to be well-posed. Numerical res
 ults demonstrate state-of-the-art surface reconstruction with consistent S
 DF gradients. The talk will also discuss an adaptation of the classical na
 rrow-band method to neural settings\, enabling accurate PDE solutions on z
 ero-level sets and opening new directions in neural geometric PDEs.\n\nhtt
 ps://math-events.uni-bonn.de/event/833/
LOCATION:Endenicher Allee 60/1-016 - Lipschitzsaal (Mathezentrum)
URL:https://math-events.uni-bonn.de/event/833/
END:VEVENT
END:VCALENDAR