ANRL-PyTorch

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

Usage

pip install -r requirements.txt

python main.py

Attention

1. We did not use exactly the same model weight initialization method as in ANRL.
2. The underlying implementation of PyTorch may result in different results from tensorflow under the same hyper-parameter settings.