An implementation of ANRL based on PyTorch. Most code of data processing in this repository comes from cszhangzhen/ANRL.
Thanks to Stonesjtu/Pytorch-NCE, We can easily achieve our aim similar to tf.nn.sampled_softmax_loss.
Original paper ANRL: Attributed Network Representation Learning via Deep Neural Networks
pip install -r requirements.txt
python main.py
- We did not use exactly the same model weight initialization method as in ANRL.
- The underlying implementation of PyTorch may result in different results from tensorflow under the same hyper-parameter settings.