NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:5875
Title:Exponential Family Estimation via Adversarial Dynamics Embedding

A novel approach to ML learning by taking the dual of the log-partition function, resulting in a saddle-point (or adversarial) problem with kinetic variables. Reviews were consistent that the approach was novel and may spur further research.