Skip to content

Latest commit

 

History

History
42 lines (31 loc) · 1.37 KB

README.md

File metadata and controls

42 lines (31 loc) · 1.37 KB

DGL & Pytorch implementation of Enhanced Graph Embedding with Side information (EGES)

Version

dgl==0.6.1, torch==1.9.0

Paper

Billion-scale Commodity Embedding for E-commerce Recommendation in Alibaba:

https://arxiv.org/pdf/1803.02349.pdf

https://arxiv.org/abs/1803.02349

Dataset

https://wx.jdcloud.com/market/jdata/list/17

How to run

Create folder named data. Download two csv files from here into the data folder.

Run command: python main.py with default configuration, and the following message will shown up:

Using backend: pytorch
Num skus: 1006, num brands: 221, num shops: 308, num cates: 55
Epoch 00000 | Step 00000 | Step Loss 0.8452 | Epoch Avg Loss: 0.8452
Evaluate link prediction AUC: 0.5557
Epoch 00001 | Step 00000 | Step Loss 0.7293 | Epoch Avg Loss: 0.7293
Evaluate link prediction AUC: 0.5774
Epoch 00002 | Step 00000 | Step Loss 0.7157 | Epoch Avg Loss: 0.7157
Evaluate link prediction AUC: 0.5764
...
Epoch 00028 | Step 00000 | Step Loss 0.7105 | Epoch Avg Loss: 0.7105
Evaluate link prediction AUC: 0.5880
Epoch 00029 | Step 00000 | Step Loss 0.7115 | Epoch Avg Loss: 0.7115
Evaluate link prediction AUC: 0.5914

The AUC of link-prediction task on test graph is computed after each epoch is done.

Reference

https://github.com/nonva/eges

https://github.com/wangzhegeek/EGES.git