BEGIN:VCALENDAR
VERSION:2.0
PRODID:researchseminars.org
CALSCALE:GREGORIAN
X-WR-CALNAME:researchseminars.org
BEGIN:VEVENT
SUMMARY:Qi Tang (Georgia Tech)
DTSTART:20260601T111500Z
DTEND:20260601T120000Z
DTSTAMP:20260417T004444Z
UID:cam/103
DESCRIPTION:Title: <a href="https://researchseminars.org/talk/cam/103/">St
 ructure-Preserving Neural Operators for Convection–Diffusion Dynamics</a
 >\nby Qi Tang (Georgia Tech) as part of CAM seminar\n\nLecture held in MV:
 L14.\n\nAbstract\nLearning convection–diffusion dynamics with neural ope
 rators is difficult because transport and dissipation act on different sca
 les\, and standard neural operators often lose stability across regimes. W
 e propose a Structure-Preserving Neural Operator that captures this transp
 ort–dissipation interplay. The method uses Strang splitting to evolve hy
 perbolic and parabolic dynamics in substeps. Convection is handled by a le
 arnable semi-Lagrangian approach that follows characteristics and embeds f
 low structure directly into the architecture\, while diffusion is treated 
 through a residual correction neural operator. Experiments on variable-coe
 fficient problems and the Vlasov–Poisson–Fokker–Planck system show i
 mproved stability\, accuracy\, and long-time performance with large time s
 teps.\n
LOCATION:https://researchseminars.org/talk/cam/103/
END:VEVENT
END:VCALENDAR