Canonical Tensor Decomposition for Knowledge Base Completion

Timothée Lacroix, Nicolas Usunier, Guillaume Obozinski
6/19/2018
stat.MLcs.AIcs.LG

Official code: facebookresearch/kbc (259 stars, last commit 2021-08-16).

Abstract

The problem of Knowledge Base Completion can be framed as a 3rd-order binary tensor completion problem. In this light, the Canonical Tensor Decomposition (CP) (Hitchcock, 1927) seems like a natural solution; however, current implementations of CP on standard Knowledge Base Completion benchmarks are lagging behind their competitors. In this work, we attempt to understand the limits of CP for knowledge base completion. First, we motivate and test a novel regularizer, based on tensor nuclear $p$-norms. Then, we present a reformulation of the problem that makes it invariant to arbitrary choices in the inclusion of predicates or their reciprocals in the dataset. These two methods combined allow us to beat the current state of the art on several datasets with a CP decomposition, and obtain even better results using the more advanced ComplEx model.

View on arXivView PDF

Code Implementations(1)

facebookresearch/kbcOfficial100%

Tools for state of the art Knowledge Base Completion.

25943Shell, PythonNov 16, 20175 years agoNOASSERTION

Similar Papers

Knowledge Graph Completion by Intermediate Variables Regularization

85% match2 repos32025

Duality-Induced Regularizer for Tensor Factorization Based Knowledge Graph Completion

84% match1 repo512020

Low-Rank Tensor Completion Based on Fractional Regularization with Ky Fan p-k Norm

82% match2026

Deterministic tensor completion with hypergraph expanders

78% match1 repo32019

Low Rank Tensor Completion via Adaptive ADMM

78% match2026

Spectral gap-based deterministic tensor completion

78% match2023

Cite this paper

@article{lacroix2018canonical,
  title  = {Canonical Tensor Decomposition for Knowledge Base Completion},
  author = {Timothée Lacroix and Nicolas Usunier and Guillaume Obozinski},
  year   = {2018},
  eprint = {1806.07297},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1806.07297}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1806.07297.svg)](https://codeofpaper.com/paper/1806.07297)

Discussion