CIEGCL

This is the PyTorch implementation for CIEGCL proposed in the paper CIEGCL: Counterfactual Intervention Enhancing Graph Contrastive Learning in Implicit Feedback

1. Running environment

We develope our codes in the following environment:

Python version 3.9.12
torch==1.13.1
numpy==1.21.5
pandas=1.3.5
tqdm==4.64.1

2. Some configurable arguments

• --topks Top k for testing recommendation performance.
• --cl_weight specifies $\lambda_S$, the regularization weight for CL loss.
• --weight_decay is $\lambda$, the L2 regularization weight.
• --temp specifies $\tau$, the temperature in CL loss.
• --static_prob is the edge dropout rate.
• --batch_size is the train batch size.
• --embedding_dim is the embedding dimension for embedding based models.
• --n_layers is the layer number for GNN models.

3. Different from the default parameter settings

• Ciao

--cl_weight 0.001

• Movielens-1M

--cl_weight 0.001 --temp 0.2

• Coat

--batch_size 64 --temp 0.5 --weight_decay 0.01