The task of graph classification is currently dominated by graph kernels, which, while powerful, scale poorly to large graphs and datasets. convolutional neural networks (CNNs) offer a very appealing alternative. However, feeding graphs to CNNs is not trivial. To address this challenge