@incollection{icml2020_3323,
abstract = {Learning graph generative models is a challenging task for deep learning and has wide applicability to a range of domains like chemistry, biology and social science. However current deep neural methods suffer from limited scalability: for a graph with \textdollar n\textdollar nodes and \textdollar m\textdollar edges, existing deep neural methods require \textdollar \textbackslash Omega(n\^{}2)\textdollar complexity by building up the adjacency matrix. On the other hand, many real world graphs are actually sparse in the sense that \textdollar m\textbackslash ll n\^{}2\textdollar . Based on this, we develop a novel autoregressive model, named BiGG, that utilizes this sparsity to avoid generating the full adjacency matrix, and importantly reduces the graph generation time complexity to \textdollar O((n + m)\textbackslash log n)\textdollar . Furthermore, during training this autoregressive model can be parallelized with \textdollar O(\textbackslash log n)\textdollar synchronization stages, which makes it much more efficient than other autoregressive models that require \textdollar \textbackslash Omega(n)\textdollar . Experiments on several benchmarks show that the proposed approach not only scales to orders of magnitude larger graphs than previously possible with deep autoregressive graph generative models, but also yields better graph generation quality.},
author = {Dai, Hanjun and Nazi, Azade and Li, Yujia and Dai, Bo and Schuurmans, Dale},
booktitle = {Proceedings of Machine Learning and Systems 2020},
pages = {6116--6126},
title = {Scalable Deep Generative Modeling for Sparse Graphs},
year = {2020}
}