funsor icon indicating copy to clipboard operation
funsor copied to clipboard

Implement an expectation propagation interpretation

Open fritzo opened this issue 6 years ago • 1 comments

One approach to approximate inference is expectation propagation. Can we implement EP as an interpretation?

fritzo avatar Mar 21 '19 16:03 fritzo

Kevin Murphy recommended a paper by the ForneyLab authors

In the case where discretes are children of Gaussians (eg Gaussian state space models which generate discrete observations), you have to use EP-type methods for the local message passing.

M. Cox and B. de Vries, “Robust Expectation Propagation in Factor Graphs Involving Both Continuous and Binary Variables,” in 26th European Signal Processing Conference, 2018 [Online]. Available: https://biaslab.github.io/pdf/eusipco2018/cox_robust_expectation_propagation.pdf

fritzo avatar Apr 11 '19 20:04 fritzo