Code for paper PairRE: Knowledge Graph Embeddings via Paired Relation Vectors.

Last update: Dec 19, 2022

Overview

PairRE

Code for paper PairRE: Knowledge Graph Embeddings via Paired Relation Vectors.

This implementation of PairRE for Open Graph Benchmak datasets (ogbl-wikikg and ogbl-biokg) is based on OGB. Thanks for their contributions.

Setup

To run the code, you need the following dependencies:

Pytorch 1.4.0
ogb 1.2.2 for wikikg
ogb 1.2.4 for wikikg2

Results

The results of PairRE on ogbl-wikikg, ogbl-biokg and ogbl-wikikg2 are as follows.

ogbl-wikikg

	#Dim	#Parameters	Hardware	Test MRR	Valid MRR
PairRE	100	250,167,400	16GB GPU	0.4912 ± 0.004	0.5013 ± 0.004
PairRE	200	500,334,800	16GB GPU	0.5289 ± 0.003	0.5529 ± 0.001

ogbl-biokg

	#Dim	#Parameters	Hardware	Test MRR	Valid MRR
PairRE	2000	187,750,000	16GB GPU	0.8164 ± 0.0005	0.8172 ± 0.0005

ogbl-wikikg2

	#Dim	#Parameters	Hardware	Test MRR	Valid MRR
PairRE	100	250,167,400	16GB GPU	0.4849 ± 0.003	0.4941 ± 0.004
PairRE	200	500,334,800	16GB GPU	0.5208 ± 0.003	0.5423 ± 0.002

Running the code

ogbl-wikikg

cd wikikg && sh examples.sh

ogbl-biokg

cd biokg && sh examples.sh

ogbl-wikikg2

Please update ogb package to version 1.2.4. The hyperparameters are same to the experiments in ogbl-wikikg.

cd wikikg && sh examples.sh

The details of the optional hyperparameters can be found in examples.sh.

Code for paper PairRE: Knowledge Graph Embeddings via Paired Relation Vectors.

Related tags

Overview

PairRE

Setup

Results

ogbl-wikikg

ogbl-biokg

ogbl-wikikg2

Running the code

ogbl-wikikg

ogbl-biokg

ogbl-wikikg2

Owner

Alipay

Transfer Reinforcement Learning for Differing Action Spaces via Q-Network Representations

Detectron2-FC a fast construction platform of neural network algorithm based on detectron2

Voice assistant - Voice assistant with python

An efficient PyTorch implementation of the evaluation metrics in recommender systems.

2.86% and 15.85% on CIFAR-10 and CIFAR-100

HybVIO visual-inertial odometry and SLAM system

A neuroanatomy-based augmented reality experience powered by computer vision. Features 3D visuals of the Atlas Brain Map slices.

quantize aware training package for NCNN on pytorch

Exploring the link between uncertainty estimates obtained via "exact" Bayesian inference and out-of-distribution (OOD) detection.

CVPR2021 Content-Aware GAN Compression

Pytorch implementation of FlowNet by Dosovitskiy et al.

A cross-document event and entity coreference resolution system, trained and evaluated on the ECB+ corpus.

Continual Learning of Long Topic Sequences in Neural Information Retrieval

Portfolio analytics for quants, written in Python

Library to enable Bayesian active learning in your research or labeling work.

Implementation of CVPR'21: RfD-Net: Point Scene Understanding by Semantic Instance Reconstruction

Code for BMVC2021 "MOS: A Low Latency and Lightweight Framework for Face Detection, Landmark Localization, and Head Pose Estimation"

This project is based on RIFE and aims to make RIFE more practical for users by adding various features and design new models

A fast, dataset-agnostic, deep visual search engine for digital art history

The Ludii general game system, developed as part of the ERC-funded Digital Ludeme Project.