scikit-learn-compatible neural network models in implemented in TensorFlow

Installation with pip is recommended:

pip install muffnn

You can install the dependencies via:

pip install -r requirements.txt

If you have trouble installing TensorFlow, see this page for more details.

For development, a few additional dependencies are needed:

pip install -r dev-requirements.txt

Each estimator in the code follows the scikit-learn API. Thus usage follows the scikit-learn conventions:

from muffnn import MLPClassifier

X, y = load_some_data()

mlp = MLPClassifier()
mlp.fit(X, y)

X_new = load_some_unlabeled_data()
y_pred = mlp.predict(X_new)

Further, serialization of the TensorFlow graph and data is handled automatically when the object is pickled:

import pickle

with open('est.pkl', 'wb') as fp:
    pickle.dump(est, fp)

See CONTIBUTING.md for information about contributing to this project.

BSD-3

See LICENSE.txt for details.