This repository is an implementation of paper : Improving the Training of Graph Neural Networks with Consistency Regularization

Related tags

Deep LearningCRGNN
Overview

CRGNN

Paper : Improving the Training of Graph Neural Networks with Consistency Regularization

Environments

Implementing environment: GeForce RTX™ 3090 24GB (GPU)

Requirements

pytorch>=1.8.1

ogb=1.3.2

numpy=1.21.2

cogdl (latest version)

Training

GAMLP+RLU+SCR

For ogbn-products:

Params: 3335831
python pre_processing.py --num_hops 5 --dataset ogbn-products

python main.py --use-rlu --method R_GAMLP_RLU --stages 400 300 300 300 300 300 --train-num-epochs 0 0 0 0 0 0 --threshold 0.85 --input-drop 0.2 --att-drop 0.5 --label-drop 0 --pre-process --residual --dataset ogbn-products --num-runs 10 --eval 10 --act leaky_relu --batch_size 50000 --patience 300 --n-layers-1 4 --n-layers-2 4 --bns --gama 0.1 --consis --tem 0.5 --lam 0.1 --hidden 512 --ema

GAMLP+MCR

For ogbn-products:

Params: 3335831
python pre_processing.py --num_hops 5 --dataset ogbn-products

python main.py --use-rlu --method R_GAMLP_RLU --stages 800 --train-num-epochs 0 --input-drop 0.2 --att-drop 0.5 --label-drop 0 --pre-process --residual --dataset ogbn-products --num-runs 10 --eval 10 --act leaky_relu --batch_size 100000 --patience 300 --n-layers-1 4 --n-layers-2 4 --bns --gama 0.1 --tem 0.5 --lam 0.5 --ema --mean_teacher --ema_decay 0.999 --lr 0.001 --adap --gap 10 --warm_up 150 --top 0.9 --down 0.8 --kl --kl_lam 0.2 --hidden 512

GIANT-XRT+GAMLP+MCR

Please follow the instruction in GIANT to get the GIANT-XRT node features.

For ogbn-products:

Params: 2144151
python pre_processing.py --num_hops 5 --dataset ogbn-products --giant_path " "

python main.py --use-rlu --method R_GAMLP_RLU --stages 800 --train-num-epochs 0 --input-drop 0.2 --att-drop 0.5 --label-drop 0 --pre-process --residual --dataset ogbn-products --num-runs 10 --eval 10 --act leaky_relu --batch_size 100000 --patience 300 --n-layers-1 4 --n-layers-2 4 --bns --gama 0.1 --tem 0.5 --lam 0.5 --ema --mean_teacher --ema_decay 0.99 --lr 0.001 --adap --gap 10 --warm_up 150 --kl --kl_lam 0.2 --hidden 256 --down 0.7 --top 0.9 --giant

SAGN+MCR

For ogbn-products:

Params: 2179678
python pre_processing.py --num_hops 3 --dataset ogbn-products

python main.py --method SAGN --stages 1000 --train-num-epochs 0 --input-drop 0.2 --att-drop 0.4 --pre-process --residual --dataset ogbn-products --num-runs 10 --eval 10 --batch_size 100000 --patience 300 --tem 0.5 --lam 0.5 --ema --mean_teacher --ema_decay 0.99 --lr 0.001 --adap --gap 20 --warm_up 150 --top 0.85 --down 0.75 --kl --kl_lam 0.01 --hidden 512 --zero-inits --dropout 0.5 --num-heads 1  --label-drop 0.5  --mlp-layer 2 --num_hops 3 --label_num_hops 14

GIANT-XRT+SAGN+MCR

Please follow the instruction in GIANT to get the GIANT-XRT node features.

For ogbn-products:

Params: 1154654
python pre_processing.py --num_hops 3 --dataset ogbn-products --giant_path " "

python main.py --method SAGN --stages 1000 --train-num-epochs 0 --input-drop 0.2 --att-drop 0.4 --pre-process --residual --dataset ogbn-products --num-runs 10 --eval 10 --batch_size 50000 --patience 300 --tem 0.5 --lam 0.5 --ema --mean_teacher --ema_decay 0.99 --lr 0.001 --adap --gap 20 --warm_up 100 --top 0.85 --down 0.75 --kl --kl_lam 0.02 --hidden 256 --zero-inits --dropout 0.5 --num-heads 1  --label-drop 0.5  --mlp-layer 1 --num_hops 3 --label_num_hops 9 --giant

Use Optuna to search for C&S hyperparameters

We searched hyperparameters using Optuna on validation set.

python post_processing.py --file_name --search

GAMLP+RLU+SCR+C&S

python post_processing.py --file_name --correction_alpha 0.4780826957236622 --smoothing_alpha 0.40049734940262954

GIANT-XRT+SAGN+MCR+C&S

python post_processing.py --file_name --correction_alpha 0.42299283241438157 --smoothing_alpha 0.4294212449832242

Node Classification Results:

Performance on ogbn-products(10 runs):

Methods Validation accuracy Test accuracy
SAGN+MCR 0.9325±0.0004 0.8441±0.0005
GAMLP+MCR 0.9319±0.0003 0.8462±0.0003
GAMLP+RLU+SCR 0.9292±0.0005 0.8505±0.0009
GAMLP+RLU+SCR+C&S 0.9304±0.0005 0.8520±0.0008
GIANT-XRT+GAMLP+MCR 0.9402±0.0004 0.8591±0.0008
GIANT-XRT+SAGN+MCR 0.9389±0.0002 0.8651±0.0009
GIANT-XRT+SAGN+MCR+C&S 0.9387±0.0002 0.8673±0.0008

Citation

Our paper:

@misc{zhang2021improving,
      title={Improving the Training of Graph Neural Networks with Consistency Regularization}, 
      author={Chenhui Zhang and Yufei He and Yukuo Cen and Zhenyu Hou and Jie Tang},
      year={2021},
      eprint={2112.04319},
      archivePrefix={arXiv},
      primaryClass={cs.SI}
}

GIANT paper:

@article{chien2021node,
  title={Node Feature Extraction by Self-Supervised Multi-scale Neighborhood Prediction},
  author={Eli Chien and Wei-Cheng Chang and Cho-Jui Hsieh and Hsiang-Fu Yu and Jiong Zhang and Olgica Milenkovic and Inderjit S Dhillon},
  journal={arXiv preprint arXiv:2111.00064},
  year={2021}
}

GAMLP paper:

@article{zhang2021graph,
  title={Graph attention multi-layer perceptron},
  author={Zhang, Wentao and Yin, Ziqi and Sheng, Zeang and Ouyang, Wen and Li, Xiaosen and Tao, Yangyu and Yang, Zhi and Cui, Bin},
  journal={arXiv preprint arXiv:2108.10097},
  year={2021}
}

SAGN paper:

@article{sun2021scalable,
  title={Scalable and Adaptive Graph Neural Networks with Self-Label-Enhanced training},
  author={Sun, Chuxiong and Wu, Guoshi},
  journal={arXiv preprint arXiv:2104.09376},
  year={2021}
}

C&S paper:

@inproceedings{
huang2021combining,
title={Combining Label Propagation and Simple Models out-performs Graph Neural Networks},
author={Qian Huang and Horace He and Abhay Singh and Ser-Nam Lim and Austin Benson},
booktitle={International Conference on Learning Representations},
year={2021},
url={https://openreview.net/forum?id=8E1-f3VhX1o}
}
Owner
THUDM
Data Mining Research Group at Tsinghua University
THUDM
A Pytorch reproduction of Range Loss, which is proposed in paper 《Range Loss for Deep Face Recognition with Long-Tailed Training Data》

RangeLoss Pytorch This is a Pytorch reproduction of Range Loss, which is proposed in paper 《Range Loss for Deep Face Recognition with Long-Tailed Trai

Youzhi Gu 7 Nov 27, 2021
Scalable Graph Neural Networks for Heterogeneous Graphs

Neighbor Averaging over Relation Subgraphs (NARS) NARS is an algorithm for node classification on heterogeneous graphs, based on scalable neighbor ave

Facebook Research 67 Dec 03, 2022
Testing the Facial Emotion Recognition (FER) algorithm on animations

PegHeads-Tutorial-3 Testing the Facial Emotion Recognition (FER) algorithm on animations

PegHeads Inc 2 Jan 03, 2022
End-to-End Dense Video Captioning with Parallel Decoding (ICCV 2021)

PDVC Official implementation for End-to-End Dense Video Captioning with Parallel Decoding (ICCV 2021) [paper] [valse论文速递(Chinese)] This repo supports:

Teng Wang 118 Dec 16, 2022
BOOKSUM: A Collection of Datasets for Long-form Narrative Summarization

BOOKSUM: A Collection of Datasets for Long-form Narrative Summarization Authors: Wojciech Kryściński, Nazneen Rajani, Divyansh Agarwal, Caiming Xiong,

Salesforce 125 Dec 31, 2022
Contrastive Learning Inverts the Data Generating Process

Official code to reproduce the results and data presented in the paper Contrastive Learning Inverts the Data Generating Process.

71 Nov 25, 2022
A Weakly Supervised Amodal Segmenter with Boundary Uncertainty Estimation

Paper Khoi Nguyen, Sinisa Todorovic "A Weakly Supervised Amodal Segmenter with Boundary Uncertainty Estimation", accepted to ICCV 2021 Our code is mai

Khoi Nguyen 5 Aug 14, 2022
A simple version for graphfpn

GraphFPN: Graph Feature Pyramid Network for Object Detection Download graph-FPN-main.zip For training , run: python train.py For test with Graph_fpn

WorldGame 67 Dec 25, 2022
Pytorch implementation of NeurIPS 2021 paper: Geometry Processing with Neural Fields.

Geometry Processing with Neural Fields Pytorch implementation for the NeurIPS 2021 paper: Geometry Processing with Neural Fields Guandao Yang, Serge B

Guandao Yang 162 Dec 16, 2022
Pytorch library for end-to-end transformer models training and serving

Pytorch library for end-to-end transformer models training and serving

Mikhail Grankin 768 Jan 01, 2023
Repo for CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning

CReST in Tensorflow 2 Code for the paper: "CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning" by Chen Wei, Ki

Google Research 75 Nov 01, 2022
PyTorch code for the paper: FeatMatch: Feature-Based Augmentation for Semi-Supervised Learning

FeatMatch: Feature-Based Augmentation for Semi-Supervised Learning This is the PyTorch implementation of our paper: FeatMatch: Feature-Based Augmentat

43 Nov 19, 2022
Medical-Image-Triage-and-Classification-System-Based-on-COVID-19-CT-and-X-ray-Scan-Dataset

Medical-Image-Triage-and-Classification-System-Based-on-COVID-19-CT-and-X-ray-Sc

2 Dec 26, 2021
Neural machine translation between the writings of Shakespeare and modern English using TensorFlow

Shakespeare translations using TensorFlow This is an example of using the new Google's TensorFlow library on monolingual translation going from modern

Motoki Wu 245 Dec 28, 2022
Official implementation of "MetaSDF: Meta-learning Signed Distance Functions"

MetaSDF: Meta-learning Signed Distance Functions Project Page | Paper | Data Vincent Sitzmann*, Eric Ryan Chan*, Richard Tucker, Noah Snavely Gordon W

Vincent Sitzmann 100 Jan 01, 2023
YOLOX_AUDIO is an audio event detection model based on YOLOX

YOLOX_AUDIO is an audio event detection model based on YOLOX, an anchor-free version of YOLO. This repo is an implementated by PyTorch. Main goal of YOLOX_AUDIO is to detect and classify pre-defined

intflow Inc. 77 Dec 19, 2022
Video Frame Interpolation without Temporal Priors (a general method for blurry video interpolation)

Video Frame Interpolation without Temporal Priors (NeurIPS2020) [Paper] [video] How to run Prerequisites NVIDIA GPU + CUDA 9.0 + CuDNN 7.6.5 Pytorch 1

YoujianZhang 31 Sep 04, 2022
arxiv-sanity, but very lite, simply providing the core value proposition of the ability to tag arxiv papers of interest and have the program recommend similar papers.

arxiv-sanity, but very lite, simply providing the core value proposition of the ability to tag arxiv papers of interest and have the program recommend similar papers.

Andrej 671 Dec 31, 2022
Powerful unsupervised domain adaptation method for dense retrieval.

Powerful unsupervised domain adaptation method for dense retrieval

Ubiquitous Knowledge Processing Lab 191 Dec 28, 2022
Bling's Object detection tool

BriVL for Building Applications This repo is used for illustrating how to build applications by using BriVL model. This repo is re-implemented from fo

chuhaojin 47 Nov 01, 2022