AdaptiveGCL

Update: Please go to https://github.com/HKUDS/AdaGCL to view the latest relevant source code and datasets of AdaGCL, this repository is no longer valid.

This is the PyTorch implementation for AdaptiveGCL proposed in the paper Adaptive Graph Contrastive Learning for Recommendation.

1. Running environment

We develop our codes in the following environment:

• python==3.9.13
• numpy==1.23.1
• torch==1.11.0
• scipy==1.9.1

2. Datasets

Dataset	# User	# Item	# Interaction	Interaction Density
Last.FM	1,892	17,632	92,834	2.8 × $10^{-3}$
Yelp	42,712	26,822	182,357	1.6 × $10^{-4}$
BeerAdvocate	10,456	13,845	1,381,094	9.5 × $10^{-3}$

3. How to run the codes

• Last.FM

python Main.py --data lastfm --eps 1e-3 --gamma -0.95

• Yelp

python Main.py --data yelp --eps 1e-3 --ssl_reg 1 --ib_reg 1e-2

• BeerAdvocate

python Main.py --data beer --ib_reg 1 --eps 1e-3 --lambda0 1e-2