BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Controlling and Guiding Generative Models for Three-Dimensional Sh
 ape Optimization
DTSTART:20260707T110000
DTEND:20260707T130000
DTSTAMP:20260523T201420Z
UID:61a3a8ca01995e053d50fe53ad8ac93ac912d28af199b00f3263009d
CATEGORIES:Conferences - Seminars
DESCRIPTION:Emilien Seiler\nEDIC candidacy exam\nExam president: Prof. Mar
 tin Rajman\nThesis advisor: Prof. Pascal Fua\nCo-examiner: Prof. Mark Paul
 y\n\nAbstract\nOptimizing 3D shapes within the latent spaces of deep gener
 ative models is fundamental to computer assisted engineering\, yet remains
  prone to a critical failure mode we term manifold drift: the tendency of 
 gradient-based optimization to move latent vectors away from the manifold 
 of valid shapes. This problem is exacerbated in state-of-the-art 3D shape 
 generative models that operate in increasingly high-dimensional latent spa
 ces where valid shapes occupy a vanishingly small fraction of the full spa
 ce. Existing mitigation strategies\, including latent regularization and f
 low-matching approaches\, either sacrifice expressiveness\, demand a diffi
 cult trade-off between objective guidance and generative fidelity that rem
 ains prone to manifold drift\, or are computationally infeasible to scale 
 to modern\, large-capacity 3D shape models. We introduce a novel optimizer
 -corrector framework that alternates between gradient steps for objective 
 minimization and guided flow matching to drive the latent state back to th
 e valid shape manifold. By decoupling objective minimization from flow-bas
 ed correction\, optimizing freely and correcting strictly\, this alternati
 ng design avoids inherent trade-offs\, preserving geometric validity witho
 ut sacrificing expressiveness while remaining computationally feasible on 
 modern 3D shape models. We demonstrate its effectiveness across generative
  priors of varying complexity\, from simple vector latent spaces to large-
 scale architectures across a variety of downstream optimization tasks\, in
 cluding aerodynamic drag reduction and object compliance optimization.\n\n
 Selected papers\n 
LOCATION:BC 333 https://plan.epfl.ch/?room==BC%20333
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR