Anonymous Code for paper "Node Number Awareness Representation for Graph Similarity Learning"

Introduction

This repo contains the source code and dataset for our paper Node Number Awareness Representation for Graph Similarity Learning. The main directory contains the code files for N2AGim and the landmark file contains the code files for GSL2. We provide the trained model parameters for N2AGim in the ./result/ and ./result_norm/ files and for GSL2 in the ./landmark/results/.

Requirements

See "requirements.txt". Please following the https://pytorch-geometric.readthedocs.io/en/1.7.1/notes/installation.html to install the torch_geometric==1.7.1.

Evaluation

The file structure of the ./result/ for N2AGim is

├── AIDS700nef
│   └── N2Gim
│       └── sum_none_pGinFalse_woNodeattFalse_woGraphattFalse
│           └── 2022-09-10_12-32-16
│               ├── final_0.pt
│               ├── final_optim_0.pt
│               ├── minvaild_0.pt
│               └── minvaild_optim_0.pt
├── IMDBMulti
│   └── N2Gim
│       └── sum_none_pGinFalse_woNodeattFalse_woGraphattFalse
│           └── 2022-09-10_16-48-56
│               ├── final_0.pt
│               ├── final_optim_0.pt
│               ├── minvaild_0.pt
│               └── minvaild_optim_0.pt
└── LINUX
    └── N2Gim
        └── sum_none_pGinFalse_woNodeattFalse_woGraphattFalse
            └── 2022-09-10_13-49-27
                ├── final_0.pt
                ├── final_optim_0.pt
                ├── minvaild_0.pt
                └── minvaild_optim_0.pt

You can evaluate our N2AGim by using the following instructions:

python test.py --file_name result/AIDS700nef/N2Gim/sum_none_pGinFalse_woNodeattFalse_woGraphattFalse/2022-09-10_12-32-16/

python test.py --file_name result/LINUX/N2Gim/sum_none_pGinFalse_woNodeattFalse_woGraphattFalse/2022-09-10_13-49-27/

python test.py --file_name result/IMDBMulti/N2Gim/sum_none_pGinFalse_woNodeattFalse_woGraphattFalse/2022-09-10_16-48-56/

You can evaluate our GSL2 by using the following instructions:

cd landmark

python mlp_test.py --dataset AIDS700nef --size 60

python mlp_test.py --dataset LINUX --size 30

python mlp_test.py --dataset IMDBMulti --size 70

Run our code

To train our N2AGim :

python train.py --datasets AIDS700nef --model N2Gim --graph_level_pooling sum

python train.py --datasets LINUX --model N2Gim --graph_level_pooling sum

python train.py --datasets IMDBMulti --model N2Gim --graph_level_pooling sum

To train our N2AGim with ATS2 similarity metric :

python train_norm.py --datasets AIDS700nef --model N2Gim --graph_level_pooling sum

python train_norm.py --datasets LINUX --model N2Gim --graph_level_pooling sum

python train_norm.py --datasets IMDBMulti --model N2Gim --graph_level_pooling sum

To train our GSL2 after training the N2AGim with ATS2 similarity metric:

cd landmark 

python mlp.py --dataset AIDS700nef --size 60

python mlp.py --dataset LINUX --size 30

python mlp.py --dataset IMDBMulti --size 70