We study the problem of node classification on graphs with few-shot novel labels, which has two distinctive properties: (1) There are novel labels to emerge in the graph; (2) The novel labels have only a few representative nodes for training a classifier. The study of this problem is instructive and corresponds to many applications such as recommendations for newly formed groups with only a few users in online social networks. To cope with this problem, we propose a novel Meta Transformed Network Embedding framework (MetaTNE), which consists of three modules: (1) A \emph{structural module} provides each node a latent representation according to the graph structure.

The overall novelty of the proposed model is limited to some extend. I think this module is very similar to meta-GNN. Both of them conduct adaptation on the support set, then do evaluation on the query set, though they employ prototype and MAML respectively. In my view, the overall model stands on the shoulder on some traditional approaches, and seems a bit incremental. Could some other approaches, such as fine-tune (which is often utilized as the comparison with meta-learning), solve this novel label problem?

The paper studies an interesting problem and formulation which can be used for a few-shot classification and cool-start recommendation. The authors provided a novel transformation function and a training scheduler in the MAML framework which the reviewers appreciated. There was some concerns initially but the rebuttal did clarify some of the confusion. In the end, reviewers were convinced that paper offers some novel ideas and it will be a great baseline for few-shot node classification methods later.

We study the problem of node classification on graphs with few-shot novel labels, which has two distinctive properties: (1) There are novel labels to emerge in the graph; (2) The novel labels have only a few representative nodes for training a classifier. The study of this problem is instructive and corresponds to many applications such as recommendations for newly formed groups with only a few users in online social networks. To cope with this problem, we propose a novel Meta Transformed Network Embedding framework (MetaTNE), which consists of three modules: (1) A \emph{structural module} provides each node a latent representation according to the graph structure. Additionally, we introduce an \emph{embedding transformation function} that remedies the deficiency of the straightforward use of meta-learning. Inherently, the meta-learned prior knowledge can be used to facilitate the learning of few-shot novel labels.