Wayfear / FBNETGEN

FBNETGEN

Dataset

PNC and ABCD

PNC can be accessed from NIH, and ABCD can be accessed from NIMH Data Archive.

ABIDE

For those who can not access ABCD and PNC dataset, we also provide an open-source dataset, ABIDE. Please follow the instruction to download and process this dataset.

Usage

PNC

python main.py --config_filename setting/pnc_fbnetgen.yaml

ABCD

python main.py --config_filename setting/abcd_fbnetgen.yaml

ABIDE

python main.py --config_filename setting/abide_fbnetgen.yaml

Hyper parameters

All hyper parameters can be tuned in setting files.

model:
  # For the model type, there are 3 choices: "seq", "gnn" or "fbnetgen". 
  type: fbnetgen

  # For the feature extractor, there are 2 choices: "gru" or "cnn".
  extractor_type: gru

  # For the feature extractor, there are 2 choices: "product" or "linear". 
  # We suggest to use "product", since it is faster.
  graph_generation: product

  # Two hyperparameters are tuned in our paper.
  embedding_size: 8
  window_size: 8



train:
  # For training method, there are 2 choices: "normal" or "bilevel".
  # "bilevel" will be in effect only if the model.type is set as "fbnetgen"
  # We suggest to use "normal".
  method: normal
  
  # If the model.type is set as "gnn", this hyper parameter will be in effect.
  # There are 2 choices: "uniform" or "pearson".
  pure_gnn_graph: pearson

Performance