Awesome

NICKI-NICKI(hide)

Parts of the code are from https://github.com/tkipf/pygcn. <br /> Our datasets can be found here.

The code is set for training NICKI for cora dataset with base class 5 and target class 0. <br />

Steps to run the pipeline

• For running NICKI(hide) -

  • Uncomment first line in script.sh
  • Put value: HIDDEN=True in main_new.py

• For changing base class-

  • Change value of base_class variable in main_new.py and feature_generator_vae.py

• For changing target class-

  • Change value 359 in line: num_fake_nodes = int(budget_coeff*359) to number of target nodes in test set, in main_new.py
  • Change num_of_target_nodes = 359 value also in feature_generator_vae.py
  • Change parameter -target to the designated target class in script.sh 'currently for class 0: -target 0'

• For budget change-

  • Update first line of script.sh with required budget, currently 0.03

• For running other datasets-

  • Datasets are available in dataset folder, refer to paper details for changes in code.

• Run the code using command - sh script.sh