am2

Stuff made on the machine learning course at my university

Simple neuron

Instantiating a neuron with a sigmoid transfer function:

from am2 import Neuron, SigmoidTransferFunction

neuron = Neuron(SigmoidTransferFunction, weights=[0, 1, 1])
print neuron.run([3, 2, 1])

Or with a step transfer function:

from am2 import Neuron, StepTransferFunction

neuron = Neuron(StepTransferFunction, function=lambda y: y > 2, weights=[0, 1, 1])
print neuron.run([1, 1, 1])

function is optional (default is lambda y: y >= 1)

Perceptron

A simple Perceptron is implemented:

from am2 import Perceptron

train_dataset = [
    ((1, 0, 0), 1),
    ((1, 0, 1), 1),
    ((1, 1, 0), 1),


    ((1, 1, 1), 0),
]

perceptron = Perceptron(function=lambda y: y >= 1)
perceptron.train(train_dataset)  # executa o algoritmo de treinamento

print perceptron.run((1, 0, 0))

matplotlib is used to plot graphs (you must have freetype and libpng installed)