Deep

Deep provides a scikit-learn interface to deep learning algorithms.

Supported Architectures

• Tied Autoencoder
• Untied Autoencoder
• Denoising Autoencoder

Init and Fit

Here is a simple example of fitting an autoencoder on MNIST:

from deep.autoencoder import TiedAE
from deep.datasets import load_mnist

X = load_mnist()[0][0]
ae = TiedAE(100)
ae.fit(X)

This example initializes a tied weight autoencoder with 100 hidden sigmoid units.

Installation

Automatic installation:

pip install deep

Deep is listed in PyPI and can be installed with pip or easy_install.