Interstellar

The code for our paper "Interstellar: Searching Recurrent Architecture for Knowledge Graph Embedding" in NeurIPS 2020.

Instructions

For the sake of ease, a quick instruction is given for readers to reproduce the searching process. Note that the programs are tested on Linux (Red Hat 4.8.5-39), Python 3.7.6 from Anaconda 4.8.5.

For data packages, please unpack the data.zip file.

Install required packages

pip install -r requirements

Interstellar searching on entity alignment

python -W ignore train_align.py

Interstellar searching on link prediction

python -W ignore train_link.py

For reproducing the results in our tables, please refer to the configurations in "run.sh". Feel free to change the architectures and hyper-parameters for your customized evaluations.

bash run.sh