BCN

This package contains an implementation of Boosted Configuration (*neural*) Networks (BCNs). How do BCNs work? By creating ensembles (boosting in a supervised way) of single-layered feedforward (neural) Networks. Interested in learning more about BCNs? You can read:

• https://www.researchgate.net/publication/380760578_Boosted_Configuration_neural_Networks_for_supervised_classification
• https://thierrymoudiki.github.io/blog/2022/07/21/r/misc/boosted-configuration-networks
• https://thierrymoudiki.github.io/blog/2022/10/05/python/explainableml/interpretation-and-PI-for-BCN.

It's worth mentioning that this Python package is built on top of the R package, thanks to rpy2.

Installing BCN

• the development version: pip install -U git+https://github.com/Techtonique/bcn_python.git
• the stable version: pip install BCN

Using BCN

As of 08/12/2022, the package's interface contains two classes: BCNClassifier and BCNRegressor. If you're familiar with scikit-learn, then using the package will be straightforward (you can use fit, predict, cross_val_score, GridSearchCV, etc.), as demo'ed in the following example. Otherwise, it's relatively easy to grasp.

• BCN for Classification:

  import BCN as bcn # this line takes a long time to run, ONLY the first time it's run
  import numpy as np
  from sklearn.datasets import load_breast_cancer
  from sklearn.model_selection import train_test_split
  from sklearn import metrics
  
  
  dataset = load_breast_cancer()
  X = dataset.data
  y = dataset.target
  
  # split data into training test and test set
  X_train, X_test, y_train, y_test = train_test_split(X, y,
                                                  test_size=0.2, random_state=123)
  
  clf = bcn.BCNClassifier(**{'B': 237,
  'activation': 'sigmoid',
  'col_sample': 0.9628819950928769,
  'lam': 0.6163764764960539,
  'nu': 0.6683372199442862,
  'r': 0.9199609524470921,
  'tol': 4.8550370180201114e-06})
  
  clf.fit(X_train, y_train)
  
  preds = clf.predict(X_test)
  
  print(np.mean(y_test == preds))
  
  print(metrics.classification_report(preds, y_test))
  

• BCN for Regression

  BCN for Regression works exactly the same way as BCN for Classification. You just need to use a continuous response y as input (no integers in the response).

Credits

This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.