Recently, there has been great success in applying deep neural networks on\ngraph structured data. Most work, however, focuses on either node- or\ngraph-level supervised learning, such as node, link or graph classification or\nnode-level unsupervised learning (e.g. node clustering). Despite its wide range\nof possible applications, graph-level unsupervised learning has not received\nmuch attention yet. This might be mainly attributed to the high representation\ncomplexity of graphs, which can be represented by n! equivalent adjacency\nmatrices, where n is the number of nodes. In this work we address this issue by\nproposing a permutation-invariant variational autoencoder for graph structured\ndata. Our proposed model indirectly learns to match the node ordering of input\nand output graph, without imposing a particular node ordering or performing\nexpensive graph matching. We demonstrate the effectiveness of our proposed\nmodel on various graph reconstruction and generation tasks and evaluate the\nexpressive power of extracted representations for downstream graph-level\nclassification and regression.\n