Awesome

GraphControl: Adding Conditional Control to Universal Graph Pre-trained Models for Graph Domain Transfer Learning

Official implementation of paper <br>GraphControl: Adding Conditional Control to Universal Graph Pre-trained Models for Graph Domain Transfer Learning <br>

Yun Zhu*, Yaoke Wang*, Haizhou Shi, Zhenshuo Zhang, Dian Jiao, Siliang Tang†

In WWW 2024

Overview

This is the first work to solve the "transferability-specificity dilemma" in graph domain transfer learning. To address this challenge, we introduce an innovative deployment module coined as GraphControl, motivated by ControlNet, to realize better graph domain transfer learning. The overview of our method is depicted as:

Setup

conda create -n GraphControl python==3.9
conda activate GraphControl 
conda install pytorch==2.1.0 torchaudio==2.1.0 cudatoolkit=12.1 -c pytorch -c conda-forge
pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-2.1.0+cu121.html

Download GCC Pretrained Weight

Download GCC checkpoints Download GCC checkpoint from https://drive.google.com/file/d/1lYW_idy9PwSdPEC7j9IH5I5Hc7Qv-22-/view and save it into ./checkpoint/gcc.pth.

For Attributed Graphs

Only GCC

CUDA_VISIBLE_DEVICES=0 python gcc.py --lr 1e-3 --epochs 100 --dataset Cora_ML --model GCC --use_adj --seeds 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

GCC with GraphControl

CUDA_VISIBLE_DEVICES=0 python graphcontrol.py --dataset Cora_ML --epochs 100 --lr 0.5 --optimizer adamw --weight_decay 5e-4 --threshold 0.17 --walk_steps 256 --restart 0.8 --seeds 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

For Non-attributed Graphs

For non-attribute graphs, we need to generate nodes attributes through node2vec firstly

CUDA_VISIBLE_DEVICES=0 python node2vec.py --dataset Hindex --lr 1e-2 --epochs 100

Then, we can train it as the same way with attributed graphs

CUDA_VISIBLE_DEVICES=0 python graphcontrol.py --dataset Hindex --epochs 100 --lr 0.1 --optimizer sgd --weight_decay 5e-4 --threshold 0.17 --walk_steps 256 --restart 0.5 --seeds 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Illustration of arguements

--dataset: default Cora_ML, [Cora_ML, Photo, Physics, DBLP, usa, brazil, europe, Hindex] can also be choosen
--model: default GCC_GraphControl, [GCC, GCC_GraphControl] can also be choosen. GCC refers to utilizing GCC as a pre-trained model and fine-tuning it on target data. On the other hand, GCC_GraphControl involves incorporating GraphControl with GCC to address the "transferability-specificity dilemma." Additional pre-trained models will be introduced in the updated version.

More details and explanations are in utils/args.py