Particle Flow Bayes’ Rule
[edit]
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:10221031, 2019.
Abstract
We present a particle flow realization of Bayes’ rule, where an ODEbased neural operator is used to transport particles from a prior to its posterior after a new observation. We prove that such an ODE operator exists. Its neural parameterization can be trained in a metalearning framework, allowing this operator to reason about the effect of an individual observation on the posterior, and thus generalize across different priors, observations and to sequential Bayesian inference. We demonstrated the generalization ability of our particle flow Bayes operator in several canonical and high dimensional examples.
Related Material


