Hypernetwork Knowledge Graph Embeddings

TL;DR

This paper introduces a hypernetwork-based approach for knowledge graph link prediction that outperforms previous models like ConvE by generating relation-specific convolutional filters, combining neural networks with tensor factorization insights.

Contribution

The paper presents a novel hypernetwork architecture for knowledge graph embeddings that improves link prediction accuracy and provides a theoretical connection to tensor factorization models.

Findings

Outperforms ConvE and previous methods on standard datasets.

Frames convolutional filters as a form of tensor factorization.

Highlights the role of sparsity and parameter tying in model effectiveness.

Abstract

Knowledge graphs are graphical representations of large databases of facts, which typically suffer from incompleteness. Inferring missing relations (links) between entities (nodes) is the task of link prediction. A recent state-of-the-art approach to link prediction, ConvE, implements a convolutional neural network to extract features from concatenated subject and relation vectors. Whilst results are impressive, the method is unintuitive and poorly understood. We propose a hypernetwork architecture that generates simplified relation-specific convolutional filters that (i) outperforms ConvE and all previous approaches across standard datasets; and (ii) can be framed as tensor factorization and thus set within a well established family of factorization models for link prediction. We thus demonstrate that convolution simply offers a convenient computational means of introducing sparsity and parameter tying to find an effective trade-off between non-linear expressiveness and the number of parameters to learn.