We propose to study the problem of few shot graph classification in graph\nneural networks (GNNs) to recognize unseen classes, given limited labeled graph\nexamples. Despite several interesting GNN variants being proposed recently for\nnode and graph classification tasks, when faced with scarce labeled examples in\nthe few shot setting, these GNNs exhibit significant loss in classification\nperformance. Here, we present an approach where a probability measure is\nassigned to each graph based on the spectrum of the graphs normalized\nLaplacian. This enables us to accordingly cluster the graph base labels\nassociated with each graph into super classes, where the Lp Wasserstein\ndistance serves as our underlying distance metric. Subsequently, a super graph\nconstructed based on the super classes is then fed to our proposed GNN\nframework which exploits the latent inter class relationships made explicit by\nthe super graph to achieve better class label separation among the graphs. We\nconduct exhaustive empirical evaluations of our proposed method and show that\nit outperforms both the adaptation of state of the art graph classification\nmethods to few shot scenario and our naive baseline GNNs. Additionally, we also\nextend and study the behavior of our method to semi supervised and active\nlearning scenarios.\n