NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:1695
Title:Copula-like Variational Inference

The paper proposes a new variational family based on copula-like construction, with efficient sampling from the proposed copula-like density. The framework provides a new way of modeling multivariate dependency structure with a tradeoff between parsimony and flexibility. The reviewers are in general agreement that the technical contributions are adequate, but the authors are encouraged to analyze and explain the representative power and flexibility of the proposed variational family in greater detail and depth.