A Capsule Network-based Model for Learning Node Embeddings
This work addresses node embedding learning for graph analysis, presenting an incremental improvement with a novel capsule-based method.
The paper tackles the problem of learning low-dimensional node embeddings in graph-structured data by proposing Caps2NE, an unsupervised model using two capsule layers, which achieves state-of-the-art performance on benchmark datasets for node classification.
In this paper, we focus on learning low-dimensional embeddings for nodes in graph-structured data. To achieve this, we propose Caps2NE -- a new unsupervised embedding model leveraging a network of two capsule layers. Caps2NE induces a routing process to aggregate feature vectors of context neighbors of a given target node at the first capsule layer, then feed these features into the second capsule layer to infer a plausible embedding for the target node. Experimental results show that our proposed Caps2NE obtains state-of-the-art performances on benchmark datasets for the node classification task. Our code is available at: \url{https://github.com/daiquocnguyen/Caps2NE}.