SLAPS-GNN

This repo contains the implementation of the model proposed in SLAPS: Self-Supervision Improves Structure Learning for Graph Neural Networks.

Datasets

ogbn-arxiv dataset will be loaded automatically, while Cora, Citeseer, and Pubmed are included in the GCN package, available here. Place the relevant files in the folder data_tf.

Dependencies

Python version 3.7.2
Numpy version 1.18.5
PyTorch version 1.5.1
DGL version 0.5.2
sklearn version 0.21.3
scipy version 1.2.1
torch-geometric 1.6.1
ogb version 1.2.3

To train the models, you need a machine with a GPU.

To install the dependencies, it is recommended to use a virtual environment. You can create a virtual environment and install all the dependencies with the following command:

conda env create -f environment.yml

The file requirements.txt was written for CUDA 9.2 and Linux so you may need to adapt it to your infrastructure.

Usage

To run the model you should define the following parameters:

dataset: The dataset you want to run the model on
ntrials: number of runs
epochs_adj: number of epochs
epochs: number of epochs for GNN_C (used for knn_gcn and 2step learning of the model)
lr_adj: learning rate of GNN_DAE
lr: learning rate of GNN_C
w_decay_adj: l2 regularization parameter for GNN_DAE
w_decay: l2 regularization parameter for GNN_C
nlayers_adj: number of layers for GNN_DAE
nlayers: number of layers for GNN_C
hidden_adj: hidden size of GNN_DAE
hidden: hidden size of GNN_C
dropout1: dropout rate for GNN_DAE
dropout2: dropout rate for GNN_C
dropout_adj1: dropout rate on adjacency matrix for GNN_DAE
dropout_adj2: dropout rate on adjacency matrix for GNN_C
dropout2: dropout rate for GNN_C
k: k for knn initialization with knn
lambda_: weight of loss of GNN_DAE
nr: ratio of zeros to ones to mask out for binary features
ratio: ratio of ones to mask out for binary features and ratio of features to mask out for real values features
model: model to run (choices are end2end, knn_gcn, or 2step)
sparse: whether to make the adjacency sparse and run operations on sparse mode
gen_mode: identifies the graph generator
non_linearity: non-linearity to apply on the adjacency matrix
mlp_act: activation function to use for the mlp graph generator
mlp_h: hidden size of the mlp graph generator
noise: type of noise to add to features (mask or normal)
loss: type of GNN_DAE loss (mse or bce)
epoch_d: epochs_adj / epoch2 of the epochs will be used for training GNN_DAE
half_val_as_train: use half of validation for train to get Cora390 and Citeseer370

Reproducing the Results in the Paper

In order to reproduce the results presented in the paper, you should run the following commands:

Cora

FP

Run the following command:

python main.py -dataset cora -ntrials 10 -epochs_adj 2000 -lr 0.001 -lr_adj 0.01 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 512 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.5 -dropout_adj2 0.25 -k 30 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 0 -non_linearity elu -epoch_d 5

MLP

Run the following command:

python main.py -dataset cora -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.001 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 512 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 20 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 1 -non_linearity relu -mlp_h 1433 -mlp_act relu -epoch_d 5

MLP-D

Run the following command:

python main.py -dataset cora -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.001 -w_decay 0.05 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 512 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 15 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 2 -non_linearity relu -mlp_act relu -epoch_d 5

Citeseer

FP

Run the following command:

python main.py -dataset citeseer -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.01 -w_decay 0.05 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 1024 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.4 -dropout_adj2 0.4 -k 30 -lambda_ 1.0 -nr 1 -ratio 10 -model end2end -sparse 0 -gen_mode 0 -non_linearity elu -epoch_d 5

MLP

Run the following command:

python main.py -dataset citeseer -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.001 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 1024 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 30 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 1 -non_linearity relu -mlp_act relu -mlp_h 3703 -epoch_d 5

MLP-D

Run the following command:

python main.py -dataset citeseer -ntrials 10 -epochs_adj 2000 -lr 0.001 -lr_adj 0.01 -w_decay 0.05 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 1024 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.5 -dropout_adj2 0.5 -k 20 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 2 -non_linearity relu -mlp_act tanh -epoch_d 5

Cora390

FP

Run the following command:

python main.py -dataset cora -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.01 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 512 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 20 -lambda_ 100.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 0 -non_linearity elu -epoch_d 5 -half_val_as_train 1

MLP

Run the following command:

python main.py -dataset cora -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.001 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 512 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 20 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 1 -non_linearity relu -mlp_h 1433 -mlp_act relu -epoch_d 5 -half_val_as_train 1

MLP-D

Run the following command:

python main.py -dataset cora -ntrials 10 -epochs_adj 2000 -lr 0.001 -lr_adj 0.001 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 512 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 20 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 2 -non_linearity relu -mlp_act relu -epoch_d 5 -half_val_as_train 1

Citeseer370

FP

Run the following command:

python main.py -dataset citeseer -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.01 -w_decay 0.05 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 1024 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.5 -dropout_adj2 0.5 -k 30 -lambda_ 1.0 -nr 1 -ratio 10 -model end2end -sparse 0 -gen_mode 0 -non_linearity elu -epoch_d 5 -half_val_as_train 1

MLP

Run the following command:

python main.py -dataset citeseer -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.001 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 1024 -dropout1 0.25 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 30 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 1 -non_linearity relu -mlp_act tanh -mlp_h 3703 -epoch_d 5 -half_val_as_train 1

MLP-D

Run the following command:

python main.py -dataset citeseer -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.01 -w_decay 0.05 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 1024 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 20 -lambda_ 10.0 -nr 5 -ratio 10 -model end2end -sparse 0 -gen_mode 2 -non_linearity relu -mlp_act tanh -epoch_d 5 -half_val_as_train 1

Pubmed

MLP

Run the following command:

python main.py -dataset pubmed -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.01 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 128 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.5 -dropout_adj2 0.5 -k 15 -lambda_ 10.0 -nr 5 -ratio 20 -model end2end -gen_mode 1 -non_linearity relu -mlp_h 500 -mlp_act relu -epoch_d 5 -sparse 1

MLP-D

Run the following command:

python main.py -dataset pubmed -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.01 -w_decay 0.0005 -nlayers 2 -nlayers_adj 2 -hidden 32 -hidden_adj 128 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.25 -k 15 -lambda_ 100.0 -nr 5 -ratio 20 -model end2end -sparse 0 -gen_mode 2 -non_linearity relu -mlp_act tanh -epoch_d 5 -sparse 1

ogbn-arxiv

MLP

Run the following command:

python main.py -dataset ogbn-arxiv -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.001 -w_decay 0.0 -nlayers 2 -nlayers_adj 2 -hidden 256 -hidden_adj 256 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.25 -dropout_adj2 0.5 -k 15 -lambda_ 10.0 -nr 5 -ratio 100 -model end2end -sparse 0 -gen_mode 1 -non_linearity relu -mlp_h 128 -mlp_act relu -epoch_d 2001 -sparse 1 -loss mse -noise mask

MLP-D

Run the following command:

python main.py -dataset ogbn-arxiv -ntrials 10 -epochs_adj 2000 -lr 0.01 -lr_adj 0.001 -w_decay 0.0 -nlayers 2 -nlayers_adj 2 -hidden 256 -hidden_adj 256 -dropout1 0.5 -dropout2 0.5 -dropout_adj1 0.5 -dropout_adj2 0.25 -k 15 -lambda_ 10.0 -nr 5 -ratio 100 -model end2end -sparse 0 -gen_mode 2 -non_linearity relu -mlp_act relu -epoch_d 2001 -sparse 1 -loss mse -noise normal

Cite SLAPS

If you use this package for published work, please cite the following:

@inproceedigs{fatemi2021slaps,
  title={SLAPS: Self-Supervision Improves Structure Learning for Graph Neural Networks},
  author={Fatemi, Bahare and Asri, Layla El and Kazemi, Seyed Mehran},
  booktitle={Advances in Neural Information Processing Systems},
  year={2021}
}

PyTorch code of "SLAPS: Self-Supervision Improves Structure Learning for Graph Neural Networks"

Related tags

Overview

SLAPS-GNN

Datasets

Dependencies

Usage

Reproducing the Results in the Paper

Cora

FP

MLP

MLP-D

Citeseer

FP

MLP

MLP-D

Cora390

FP

MLP

MLP-D

Citeseer370

FP

MLP

MLP-D

Pubmed

MLP

MLP-D

ogbn-arxiv

MLP

MLP-D

Cite SLAPS

Owner

Latent Network Models to Account for Noisy, Multiply-Reported Social Network Data

[ICLR 2021 Spotlight Oral] "Undistillable: Making A Nasty Teacher That CANNOT teach students", Haoyu Ma, Tianlong Chen, Ting-Kuei Hu, Chenyu You, Xiaohui Xie, Zhangyang Wang

CPF: Learning a Contact Potential Field to Model the Hand-object Interaction

Implementation of our paper 'RESA: Recurrent Feature-Shift Aggregator for Lane Detection' in AAAI2021.

Fast image augmentation library and easy to use wrapper around other libraries. Documentation: https://albumentations.ai/docs/ Paper about library: https://www.mdpi.com/2078-2489/11/2/125

Course about deep learning for computer vision and graphics co-developed by YSDA and Skoltech.

Fre-GAN: Adversarial Frequency-consistent Audio Synthesis

CVPR2022 (Oral) - Rethinking Semantic Segmentation: A Prototype View

RARA: Zero-shot Sim2Real Visual Navigation with Following Foreground Cues

Simple reference implementation of GraphSAGE.

[CIKM 2019] Code and dataset for "Fi-GNN: Modeling Feature Interactions via Graph Neural Networks for CTR Prediction"

PyTorch Implementation of Spatially Consistent Representation Learning(SCRL)

Code release for SLIP Self-supervision meets Language-Image Pre-training

Federated Learning - Including common test models for federated learning, like CNN, Resnet18 and lstm, controlled by different parser

Attentive Implicit Representation Networks (AIR-Nets)

From this paper "SESNet: A Semantically Enhanced Siamese Network for Remote Sensing Change Detection"

Code for CVPR 2021 oral paper "Exploring Data-Efficient 3D Scene Understanding with Contrastive Scene Contexts"

JAX + dataclasses

[CVPR2022] Bridge-Prompt: Towards Ordinal Action Understanding in Instructional Videos