NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:772
Title:Conditional Structure Generation through Graph Variational Generative Adversarial Nets

The work presents an interesting advance combining VAEs and GANs for graph generation as the reviewers all agree. The rebuttal for Reviewer 2's comments argue "random initialization works just as well, although initialization via spectral embedding may slightly help with the convergence 18 in model training and the robustness of the results". Please add numbers substantiating this in the camera-ready.