Paper: RoAN: A Relation-oriented Attention Network for Temporal Knowledge Graph Completion
This repository contains the implementation of the RoAN architectures described in the paper.
Install PyTorch (>= 1.1.0) following the instuctions on the PyTorch . Our code is written in Python3.
After installing the requirements, run the following command to reproduce results for RoAN-DES:
$ python main.py -dropout 0.4 -se_prop 0.36 -beta 0.5 -neg_ratio 5 -model RoAN—DES
To reproduce the results for RoAN-DED and RoAN-DET, specify model as RoAN-DED/RoAN-DET as following.
$ python main.py -dropout 0.4 -se_prop 0.36 -beta 0.5 -model RoAN—DED
$ python main.py -dropout 0.4 -se_prop 0.36 -beta 0.5 -model RoAN—DET
We use the following public codes for baselines and hyperparameters.
| Baselines | Code |
|---|---|
| TransE | link |
| TTransE | link |
| HyTE | link |
| DE-TransE / DE-DistMult / DE-SimplE | link |
| TA-TransE / TA-DistMult | link |