We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Learning Interaction Kernels for Agent Systems on Riemannian Manifolds

Abstract: Interacting agent and particle systems are extensively used to model complex phenomena in science and engineering. We consider the problem of learning interaction kernels in these dynamical systems constrained to evolve on Riemannian manifolds from given trajectory data. The models we consider are based on interaction kernels depending on pairwise Riemannian distances between agents, with agents interacting locally along the direction of the shortest geodesic connecting them. We show that our estimators converge at a rate that is independent of the dimension of the state space, and derive bounds on the trajectory estimation error, on the manifold, between the observed and estimated dynamics. We demonstrate the performance of our estimator on two classical first order interacting systems: Opinion Dynamics and a Predator-Swarm system, with each system constrained on two prototypical manifolds, the $2$-dimensional sphere and the Poincar\'e disk model of hyperbolic space.
Subjects: Machine Learning (cs.LG); Statistics Theory (math.ST)
Cite as: arXiv:2102.00327 [cs.LG]
  (or arXiv:2102.00327v3 [cs.LG] for this version)

Submission history

From: Ming Zhong [view email]
[v1] Sat, 30 Jan 2021 22:15:50 GMT (2564kb,D)
[v2] Fri, 12 Feb 2021 21:38:19 GMT (3203kb,D)
[v3] Fri, 5 Mar 2021 18:11:36 GMT (3203kb,D)

Link back to: arXiv, form interface, contact.