HS-GCN

This is our experiment codes for the paper:

HS-GCN: Hamming Spatial Graph Convolutional Networks for Recommendation.

Environment settings

• Python 3.7
• Pytorch 1.4.0
• PyTorch Geometric 1.6.1
• Numpy 1.19.5
• Pandas 1.1.3

File specification

• data_load.py : loads the raw data in path ./raw_data, and the results are saved in path ./para.
• data_triple.py : obtains the triplets for model training, and the results are saved in path ./para.
• HSGCN_model.py : implements the model framework of HS-GCN.
• model_train.py : the training process of model.
• model_test.py : the testing process of model.

Usage

• Execution sequence

  The execution sequence of codes is as follows: data_load.py--->data_triple.py--->model_train.py--->model_test.py

• Execution results

  During the execution of file model_train.py, the epoch, iteration, and training loss will be printed as the training process:

  [1,   600] loss: 1.21214
  [1,  1200] loss: 1.19586
  [1,  1800] loss: 1.18090
  [2,   600] loss: 1.13528
  [2,  1200] loss: 1.12297
  [2,  1800] loss: 1.11104
  [3,   600] loss: 1.07233
  [3,  1200] loss: 1.06290
  [3,  1800] loss: 1.05153
  ...
  

  File model_test.py should be executed after the training process, and the performance of HS-GCN will be printed:

  HR@50: 0.2052; NDCG@50: 0.3081; P@50: 0.2020