Simple Neural Net

neural_net.py

import numpy as np

def normalize(vector, span=(0,1)):
    minimum, maximum = (np.min(vector), np.max(vector))
    scaling = (span[1] - span[0]) / (maximum - minimum)
    return ((vector - minimum) * scaling) + span[0]

class NeuralNet(object):
    
    def __init__(self, layer_sizes=(10,50,20), threshold=0.5):
        self.threshold = threshold
        self.weight_matrices = []
        for i, element in enumerate(layer_sizes[:-1]):
            # Just initializing random weight matrices
            self.weight_matrices.append(np.random.rand(layer_sizes[i], layer_sizes[i+1]))

    def transfer(self, input_vector, weight_matrix):
        return normalize(np.dot(input_vector, weight_matrix)) > self.threshold

    def predict(self, input_vector):
        return reduce(self.transfer, [input_vector] + self.weight_matrices)

test = NeuralNet()
print test.predict(np.random.rand(1, 10))