Gated Graph Sequence Neural Networks
This is the code for our ICLR'16 paper:
- Yujia Li, Daniel Tarlow, Marc Brockschmidt, Richard Zemel. Gated Graph Sequence Neural Networks. International Conference on Learning Representations, 2016.
Please cite the above paper if you use our code.
The code is released under the MIT license.
th test.lua to test all the modules in the ggnn and rnn libraries.
Reproducing the bAbI tasks and graph algorithms experiment results
To run the bAbI experiments, and experiments on the two extra sequence tasks:
- Go into
bash get_10_fold_data.shto get 10 folds of bAbI data for 5 tasks (4, 15, 16, 18, 19) and do some preprocessing.
- Go into
bash generate_10_fold_data.shto get 10 folds of data for the two extra sequence tasks.
- Go back to
run_experiments.pyto run the GGNN/GGS-NN experiments, e.g.
python run_experiments.py babi18runs GGNN on bAbI task 18 for all 10 folds of data.
run_rnn_baselines.pyto run RNN/LSTM baseline experiments, e.g.
python run_rnn_baselines.py babi18 lstmruns LSTM on bAbI task 18 for all 10 folds of data.
- Make sure
"./?/init.lua"are on your lua path. For example by
- The experiment results may differ slightly from what we reported in the paper, as the datasets are randomly generated and will be different from run to run.