Hyperbolic Graph Embedding with Enhanced Semi-Implicit Variational Inference

Hyperbolic Graph Embedding with Enhanced Semi-Implicit Variational Inference in Pytorch

Hierarchical Input	Latent Output

How to run

Requirements

• NumPy
• PyTorch 1.6
• Scipy
• networkx
• json

pip install -r requirements.txt

Installing

1. Clone the repository:

   $ git clone https://github.com/AliLotfi92/ESI_HGE
   $ cd esihge

2. Install requirements:

   $ pip install -r requirements.txt

---

Arguments:

• lr: learning rate for the inference network
• dropout: Dropout rate (1 - keep probability).
• epochs: number of epochs to train the model.
• c: constant negative curvature
• K: semi-implicit vi hyperparameters
• J: semi-implicit vi hyperparameters
• dataset-str: synthetic, cora, citeseer, or pubmed

Results:

• Latent codes discoveries