Need help with ggnn?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.


Gated Graph Sequence Neural Networks

225 Stars 53 Forks MIT License 6 Commits 4 Opened issues

Services available

Need anything else?

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:

  1. Go into
    , run
    to get 10 folds of bAbI data for 5 tasks (4, 15, 16, 18, 19) and do some preprocessing.
  2. Go into
    , run
    to get 10 folds of data for the two extra sequence tasks.
  3. Go back to
    and use
    to run the GGNN/GGS-NN experiments, e.g.
    python babi18
    runs GGNN on bAbI task 18 for all 10 folds of data.
  4. Use
    to run RNN/LSTM baseline experiments, e.g.
    python babi18 lstm
    runs LSTM on bAbI task 18 for all 10 folds of data.


  • Make sure
    are on your lua path. For example by
    export LUA_PATH="./?.lua;./?/init.lua;$LUA_PATH"
  • 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.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.