NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:4594
Title:Energy-Inspired Models: Learning with Sampler-Induced Distributions

The reviewers had an in depth discussion about this submission, suggesting a weak accept based on the paper's didactic contributions, but also commenting extensively on the limitations of the paper. I've decided to recommend to accept this submission, but please incorporate the reviewers' suggested improvements into your camera ready version of the paper, and in particular: * the reviewers' request for a practical discussion was ignored * demonstrate practical implications with the unified formulation (eg "A simple extension could be [..] we leave exploring this to future work") * demonstrate briefly useful extensions such as suggesting a better auxiliary variable distribution for IWAE (which are only mentioned as a possibility like in Line 108)