Semi-supervised Network Embedding with Differentiable Deep Quantisation

TL;DR

This paper introduces d-SNEQ, a differentiable deep quantisation method for network embeddings that significantly reduces storage needs and improves performance on various network analysis tasks.

Contribution

The paper presents a novel semi-supervised network embedding approach with differentiable deep quantisation and a new evaluation metric for high-order information preservation.

Findings

d-SNEQ outperforms state-of-the-art methods in link prediction.

It achieves better results in path prediction, node classification, and node recommendation.

The method is more space- and time-efficient.

Abstract

Learning accurate low-dimensional embeddings for a network is a crucial task as it facilitates many downstream network analytics tasks. For large networks, the trained embeddings often require a significant amount of space to store, making storage and processing a challenge. Building on our previous work on semi-supervised network embedding, we develop d-SNEQ, a differentiable DNN-based quantisation method for network embedding. d-SNEQ incorporates a rank loss to equip the learned quantisation codes with rich high-order information and is able to substantially compress the size of trained embeddings, thus reducing storage footprint and accelerating retrieval speed. We also propose a new evaluation metric, path prediction, to fairly and more directly evaluate model performance on the preservation of high-order information. Our evaluation on four real-world networks of diverse characteristics shows that d-SNEQ outperforms a number of state-of-the-art embedding methods in link prediction, path prediction, node classification, and node recommendation while being far more space- and time-efficient.