Know-Evolve

This repository holds code for ICML '17 paper: "Know-Evolve: Deep Temporal Reasoning for Dynamic Knowledge Graphs"

Install

To get source code, run:

git clone https://github.com/rstriv/Know-Evolve.git

There are two steps required for complete installation:

1. Install base graphnn library:

cd code/graphnn_base

Please follow the intallation instructions provided on Readme page. This is an obsolete and slightly modified version of graphnn library.

2. Build Know-Evolve Code:

cd code/know_evolve
make

This would create a build directory. If you get any error, please check to see that paths in your Makefile are correct.

Run Experiments

To run experiments on small sample dataset (500 entities):

cd code/know_evolve
./run_small.sh

To run experiments on full dataset (will require longer testing time):

cd code/know_evolve
./run_large.sh

You can change various hyper-parameters and try your own dataset using configurations in these files.

For any questions, please contact: rstrivedi AT gatech DOT edu

Reference

@InProceedings{trivedi2017knowevolve,
  title =    {Know-Evolve: Deep Temporal Reasoning for Dynamic Knowledge Graphs},
  author =   {Trivedi, Rakshit and Dai, Hanjun and Wang, Yichen and Song, Le},
  booktitle =    {Proceedings of the 34th International Conference on Machine Learning},
  year =     {2017}
}