zhuyf8899 / ApeGNN

Code for ApeGNN

ApeGNN: Adaptive Aggregation and Pooling in GNNs for Recommendation

This is our PyTorch implementation for the paper:

run

python main.py --dataset Amazon --context_hops 3 --gpu_id 0 --gnn ApeGNN --pool sum --t_u 1 --t_i 3

Environment Requirement

The code has been tested running under Python 3.7.6. The required packages are as follows:

• pytorch == 1.7.0
• numpy == 1.18.2
• scipy == 1.4.1
• sklearn == 0.24.1
• prettytable == 2.1.0

Dataset

We use three processed datasets: AMiner, Gowalla, Yelp2018 and Amazon.

Dataset	#Users	#Items	#Interactions	Density
AMiner	5,340	14,967	163,084	0.00204
Gowalla	29,858	40,981	1,027,370	0.00084
Yelp2018	31,668	38,048	1,561,406	0.00130
Amazon	192,403	63,001	1,689,188	0.00014
ml-1m	6040	3416	999,611	0.04362

ToDo:

Exps:

1.add ml-1m: user << item

2.add attention-based baselines

✅3.u and t is learnable

4.set different t in different layer

✅5.grid search

Discussion

1. Attention-based GAT

2.