davidmcdonald1993/heat

Reference implementation of the HEAT algorithm described in https://link.springer.com/chapter/10.1007/978-3-030-62362-3_4

HEAT: Hyperbolic Embedding of Attributed Networks

Reference implementation of algorithm described in https://link.springer.com/chapter/10.1007/978-3-030-62362-3_4

Authors: David MCDONALD ([email protected]) and Shan HE ([email protected])

Requirements:

• Python3
• Numpy
• Scipy
• Scikit-learn
• Scikit-multilearn
• Keras

Required packages (pip)

Install the required packages with

pip install -r requirements.txt 

Setup environment (conda)

The conda environment is described in heat_env.yml.

Run

conda env create -f heat_env.yml

to create the environment, and

conda activate heat-env

to activate it.

How to use:

Run the code with

python main.py --edgelist path/to/edgelist.tsv --features path/to/features.csv -e *num_epochs* --embedding path/to/save/embedding --alpha *alpha_value*

Additional options can be viewed with

python main.py --help

Input Data Format

Graph

Graphs are given as tab-separated edgelists in the form:

u\tv\tw

• where w is the weight of the connection between nodes u and v
• u and v must be integers
• every int in [0, N-1] must appear in the edgelist

Node attributes and labels

labels and features should be comma separated tables indexed by node id

Citation:

If you find this useful, please use the following citation

@inproceedings{mcdonald2020heat,
  title={HEAT: Hyperbolic Embedding of Attributed Networks},
  author={McDonald, David and He, Shan},
  booktitle={International Conference on Intelligent Data Engineering and Automated Learning},
  pages={28--40},
  year={2020},
  organization={Springer}
}

heat's Issues

About generating the datasets

Hi, I am curious about how do you generate the feats and labels for these datasets. Do you mind sharing the preprocessing part of these datasets?

Also, there are two datasets (mit and pubmed) which didn't been mentioned in your paper, I am wondering where are they from? Thank you!

License

Can you add a license to the repo?

All the label is zero?

Hi I am very interested in your project. In your main.py, train_y = np.zeros([len(train_x), 1, 1], dtype=np.int32 ) , the train_y is the label right? why the label for all the positive and negative label are all 0? Thanks!