Skip to content

contribute.md

Latest commit

 

History

History
32 lines (23 loc) · 2.2 KB

contribute.md

File metadata and controls

32 lines (23 loc) · 2.2 KB

Contribute to PyGDebias

We greatly appreciate your support and contributions. Please follow the steps below to contribute to Algorithms or Datasets. For the mechani

Contribute to Datasets

Provide your datasets at any possible formats, while a dataloading() function which processes your datasets and outputs with the same formats as defined below. The best dataset split rate would be 0.5/0.25/0.25.

  • adj (Torch.Tensor): The adjacent matrix which is represented by a sparse tensor. # (node_num，node_num)
  • features (Torch.Tensor): The features of all nodes. # (node_num, feature_num)
  • labels (Torch.Tensor): lables for all nodes. # (node_num)
  • idx_train (Torch.Tensor): the index of nodes in train dataset. # (train_size)
  • idx_val (Torch.Tensor): the index of nodes in validation dataset. # (val_size)
  • idx_test (Torch.Tensor): the index of nodes in test dataset. # (test_size)
  • sens (Torch.Tensor): the vector of the sensitive group.# (node_num)
  • sens_idx (int): the index of the sensitive group. # (int)

Contribute to Algorithms

Provide your algorithm/model in a python class whose name should be exactly the model name. Your algorithm/model should at least contain two fucntions, ie,

  • fit (model specified args restricted to the outputs of the dataloading() function stated above) --> (none): execute the training process for the initiated graph mining algorithm.
  • predict (none) --> (any returns): evaluate the trained graph mining algorithm on the test set.This function should print as many metrics as possible.

For each algorithm, you should provide a description of the model, a clarification of the fit() function including the inputs types and formats, and a clarification of the predict() function including printed metrics' types and return type, better under each function definition with the format of """xxx""".

Note

When pulling requests

  • Add your new dataset into the dataset folder.
  • Add your dataloading() function with an independent python file in the main directory.
  • Add your algorithms in the PyGDebias folder.

If you are new to contributing, check the tutorial.