Variational Formulation of the Particle Flow Particle Filter
By: Yinzhuang Yi, Jorge Cortés, Nikolay Atanasov
Potential Business Impact:
Makes computer guesses better by learning from mistakes.
This paper provides a formulation of the particle flow particle filter from the perspective of variational inference. We show that the transient density used to derive the particle flow particle filter follows a time-scaled trajectory of the Fisher-Rao gradient flow in the space of probability densities. The Fisher-Rao gradient flow is obtained as a continuous-time algorithm for variational inference, minimizing the Kullback-Leibler divergence between a variational density and the true posterior density.
Similar Papers
Advances in Particle Flow Filters with Taylor Expansion Series
Computational Engineering, Finance, and Science
Improves tracking by moving "guess points" smarter.
An optimal-transport finite-particle method for driven mass diffusion
Numerical Analysis
Moves tiny things around boundaries accurately.
Hamiltonian Normalizing Flows as kinetic PDE solvers: application to the 1D Vlasov-Poisson Equations
Machine Learning (CS)
Predicts how tiny particles move and change.