TuckER: Tensor Factorization for Knowledge Graph Completion

Anonymous

Keywords: tucker decomposition, link prediction, knowledge graphs, multi-task learning

Abstract: Knowledge graphs are structured representations of real world facts. However, they typically contain only a small subset of all possible facts. Link prediction is the task of inferring missing facts based on existing ones. We propose TuckER, a relatively simple yet powerful linear model based on Tucker decomposition of the binary tensor representation of knowledge graph triples. By using this particular decomposition, parameters are shared between relations, enabling multi-task learning. TuckER outperforms previous state-of-the-art models across several standard link prediction datasets.

TL;DR: We propose TuckER, a relatively simple but powerful linear model for link prediction in knowledge graphs, based on Tucker decomposition of the binary tensor representation of knowledge graph triples.

Community Implementations: [ 4 code implementations](https://www.catalyzex.com/paper/arxiv:1901.09590/code)