Tweeted By @harvardnlp
New work with the Pyro team (https://t.co/6en944fTtd) on efficient batched discrete variable inference through tensor variable elimination.
— harvardnlp (@harvardnlp) February 13, 2019
Practically generalizes 'einsum' with batching, semirings (log-space, max, etc), and marginals. Powers Pyro's discrete directed inference. pic.twitter.com/upE6Soazkw