Neural Networks with Back-Propagation

Features

• support any number of layers and neurons in each layer
• use numpy inside to accelerate

Note: all hidden and output neurons are with tanh-type transformation

Usage

from nnet import NNet

• prepare data

accept plain list or numpy array.

• define NNet

nn = NNet([n_0, n_1, ..., n_L])

• train NNet

nn.train(X, y)

• get training MLS error

error = nn.score(X, y)

• predict

y_pred = nn.predict(X)