Ensemble Node Embeddings using Tensor Decomposition: A Case-Study on DeepWalk

TL;DR

This paper introduces TenSemble2Vec, an ensemble node embedding method that combines multiple embeddings via tensor decomposition to improve node representation quality, validated through extensive real-world data experiments.

Contribution

The paper presents a novel ensemble approach for node embeddings using tensor decomposition, effectively integrating multiple embeddings to enhance representation quality.

Findings

TenSemble2Vec outperforms individual embedding methods.

The approach effectively leverages complementary information from multiple embeddings.

Extensive experiments validate the efficiency of the proposed method.

Abstract

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.