Biobjective gradient descent for Feature Selection (BFS)

Code for BFS, a biobjective gradient descent feature selection method based on network sparsification.

Installation

To install BFS from git, clone the project using the following command:

git clone https://forge.ibisc.univ-evry.fr/tissa/BFS/tree/master/BFS.git

To reproduce the conda environment with the required packages and appropriate versions, use the following command:

For Mac users:

conda env create -f BFS_mac.yml

For Linux users:

conda env create -f BFS_linux.yml

To install BFS using pip:

python3 -m pip install BGFS

Usage

If downloaded using git, before using BFS, first go to the source files' directory using the following command:

cd BFS/source

From there or when downloaded with pip, BFS can be loaded in a python script as any other module using the import command. BFS can be used with the following commands:

from BGFS.BFS import BFS

model = BFS(n_attr, pref_vector1, pref_vector2, n_hidden, n_output)
model.compile(init_optimizer, train_optimizer, init_learning_rate, train_learning_rate, metrics)
model.fit(X_train, y_train, validation_data=(X_val, y_val))

Documentation

BFS documentation can be found on: https://bfss.readthedocs.io/

License

BFS was created by Tina Issa. It is licensed under the terms of the MIT license.