Skip to content

XiaShan1227/Self-Attention-Graph-Pooling

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

【ICML-2019 SAGPooling】Self-Attention Graph Pooling image image

1.实验参数

Parameter Value
Batch size 128
Dataset 可选: DD、MUTAG、NCI1、NCI109、PROTEINS, etc
Dropout ratio 0.5
Epochs 10000
Exp name 自命名: DD_Glo、MUTAG_Hie, etc
Gpu index 0
Hid 128
Lr 0.0005
Model 可选: SAGPooling_Global、SAGPooling_Hierarchical
Patience 40
Pooling ratio 0.5
Seed 16
Test batch size 1
Weight decay 0.0001

2.运行程序
模型:SAGPooling_Global
数据集:DD

python main.py --exp_name=DD_Glo --dataset=DD --model=SAGPooling_Global

模型:SAGPooling_Hierarchical
数据集:PROTEINS

python main.py --exp_name=PROTEINS_Hie --dataset=PROTEINS --model=SAGPooling_Hierarchical

3.实验结果(8:1:1划分数据集,只做了一次实验的准确率,保留两位小数)

DD MUTAG NCI1 NCI109 PROTEINS
SAGPooling_Global 73.11 80.00 69.10 74.40 73.21
SAGPooling_Hierarchical 67.23 70.00 66.18 70.77 69.64