An implementation of Feedforward Neural Networks for quick applications.
The quicknn is a Tensorflow-based package that aims to simplify the application of the feedforward neural networks in classification and regression problems. The main features of the quicknn package are:
- internally management of the categorical variables with one-hot-encoding(OHE) method batch-wise, just you have to feed it with pandas object;
- internally management of the validation of the data while training;
- possibility to stop the training phase, change some parameters and then resume the training from where it had remained;
- allows easy visualization of the learning curves using Tensorboard;
from quicknn import QuickNN from sklearn.datasets import load_boston from sklearn.model_selection import train_test_split X, y = load_boston(return_X_y=True) X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.25) qnn = QuickNN(list_neurons=[100, 200, 1]) qnn.fit(X_train, y_train, n_epochs=10) ## In IPython session you can stop-change-resume the training. qnn.fit(X_train, y_train, n_epochs=20) ## Just increasing the n_epochs. qnn.fit(X_train, y_train, n_epochs=30, learning_rate=0.01) ## You can change e.g., the learning_rate param while training y_pred = qnn.predict(X_val)
The dependencies are showed in requirements.txt, which can be installed with the command:
$ pip install -r requirements.txt
Then the library can easily downloaded through pip:
$ pip install quicknn
This project is licensed under the MIT License - see the LICENSE.md file for details.
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size & hash SHA256 hash help||File type||Python version||Upload date|
|quicknn-1.0.9-py3-none-any.whl (8.2 kB) Copy SHA256 hash SHA256||Wheel||py3|
|quicknn-1.0.9.tar.gz (6.5 kB) Copy SHA256 hash SHA256||Source||None|