Ensemble Node Embeddings using Tensor Decomposition: A Case-Study on DeepWalk
This work addresses the challenge of model selection in node embeddings for graph analysis, though it is incremental as it builds on existing embedding and decomposition techniques.
The authors tackled the problem of selecting optimal node embedding methods by proposing TenSemble2Vec, an ensemble approach that combines multiple embeddings using tensor decomposition, which was validated as efficient on real-world data.
Node embeddings have been attracting increasing attention during the past years. In this context, we propose a new ensemble node embedding approach, called TenSemble2Vec, by first generating multiple embeddings using the existing techniques and taking them as multiview data input of the state-of-art tensor decomposition model namely PARAFAC2 to learn the shared lower-dimensional representations of the nodes. Contrary to other embedding methods, our TenSemble2Vec takes advantage of the complementary information from different methods or the same method with different hyper-parameters, which bypasses the challenge of choosing models. Extensive tests using real-world data validates the efficiency of the proposed method.