Pynetz

A Simple Python Neural Network Library

Have you ever wanted to have the ability to play around with a simple Neural Network in Python and train it to complete tasks of your choosing? Pynetz allows you to work with Neural Networks written in Python and provides a clean and easy interface to work with.

Installation

From PyPi:

pip install pynetz

Or from source:

python setup.py install

Creating Your Own Neural Networks

import numpy as np
# initialize a Neural Network with 3 inputs at the input, a hidden layer with 100 units, and an output layer with 3 units
n = NN([3, 100, 3])
vals = np.random.random((100, 3))
X = np.split(vals, 100)  # splitting into multiple training examples
Y = np.sin(x)  # training the network to learn sin(x) as a simple example
eta = 0.1  # learning rate
epochs = 50  # max number of epochs
gen_plots = False  # don't generate plots for training error
n.train(epochs, X, Y, eta, gen_plots)

Sampling Your Neural Network

print "Expected: ", np.sin(np.array([0.8, 0.9, 0.2]))
print "Sampled: ", n.feedforward(np.array([0.8, 0.9, 0.2]))