Implementation and modification of Message Passing Neural Networks as explained in the article proposed by Gilmer et al. [1].
Requirements:
- python 3.5
- pytorch=0.1.12
- networkx=1.11
- tensorboard
- tensorboard_logger
- numpy
- joblib
Using conda create command to create a conda environment.
$ module add anaconda3/4.2.0
$ conda create -n python-3.5 python=3.5
$ source activate python-3.5
$ pip install numpy tensorboard tensorboard_logger joblib
$ conda install -c rdkit rdkit
$ conda install networkx=1.11
$ conda install pytorch=0.1.12 cuda75 -c soumith
$ git clone https://github.com/ifding/graph-neural-networks.git
$ cd graph-neural-networks
Download and convert QM9 data set:
$ python3 download_data.py qm9 -p /scratch3/feid/mpnn-data/
Train and test MPNN (default) and MPNNv2 model with GPU (default) or not:
$ python3 main.py --datasetPath /scratch3/feid/mpnn-data/qm9/dsgdb9nsd/
$ python3 main.py --datasetPath /scratch3/feid/mpnn-data/qm9/dsgdb9nsd/ --no-cuda
$ python3 main.py --datasetPath /scratch3/feid/mpnn-data/qm9/dsgdb9nsd/ --model MPNNv2
$ python3 main.py --datasetPath /scratch3/feid/mpnn-data/qm9/dsgdb9nsd/ --no-cuda --model MPNNv2
$ python3 main.py --datasetPath /scratch3/feid/mpnn-data/qm9/dsgdb9nsd/ --model MPNNv3
$ python3 main.py --datasetPath /scratch3/feid/mpnn-data/qm9/dsgdb9nsd/ --no-cuda --model MPNNv3
- [1] Gilmer et al., Neural Message Passing for Quantum Chemistry, arXiv, 2017.
- [2] Schütt, Kristof T., et al. Quantum-chemical insights from deep tensor neural networks Nature communications 8 (2017): 13890.
- [3] Duvenaud et al., Convolutional Networks on Graphs for Learning Molecular Fingerprints, NIPS, 2015.
- [4] Li et al., Gated Graph Sequence Neural Networks, ICLR, 2016.
- [5] Kipf et al., Semi-Supervised Classification with Graph Convolutional Networks, ICLR, 2017
- [6] Defferrard et al., Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering, NIPS, 2016.
- [7] Kearnes et al., Molecular Graph Convolutions: Moving Beyond Fingerprints, JCAMD, 2016.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent