home.social

#geometricdeeplearning — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #geometricdeeplearning, aggregated by home.social.

  1. Representation learning often emphasizes metric preservation. We instead build Symplectic structural invariance directly into the representation.

    arxiv.org/abs/2512.19409

    We embed Hamiltonian/symplectic geometry by making the RNN state dynamics a symplectomorphism, which preserves Legendre duality (information geometry) through time. This yields structure-preserving representations enforced by the latent dynamics, rather than imposed indirectly via the output.

    #ReservoirComputing #RepresentationLearning #InformationGeometry #SymplecticGeometry #HamiltonianDynamics #GeometricDeepLearning #DynamicalSystems #PhysicsInformedML

  2. Really looking forward to this workshop! It will be cool to discuss what equivariance can do for robotics and what the geometric dl field can learn from robotics.

    mobile.twitter.com/DianWang100

    #GeometricDeepLearning #Robotics #RSS2023

  3. On my way to #aiche 2022 in #phoenix to talk about #GeometricDeepLearning! We'll see if I can make it out of #yyz today or if my flights will continue to be delayed and cancelled 🥲