• GIF

  Visualization of a neural network learning the XOR function.

NN.py

Neural Network lib in Python

NN.py - Main Neural Network Class

Grapher.py - Data visualization Class

Data.py - Dataset Class

Graphics

graphics_pygame.py - Pygame Graphics Class

graphics_qt.py - QT Graphics Class

Install Dependencies

pip install -r requirements.txt

Example Code

from NN import *
from Data import *

data = Data()
data.generateXORData(1000, 0.2)  #generate XOR dataset with 1000 points

testData = Data()
testData.generateXORData(50, 0.2) #generate another XOR dataset with 50 points, this will be used for validation

nn = NN(2,4,3,1) #2 input neurons, 4 neurons in first hidden layer, 3 in second hidden layer and 1 in output layer
nn.activationFunction = "sigmoid" #use the sigmoid function as the activation function
nn.learningRate = 0.03

for i in range(1000): #500 epochs of training
    if(i%10 == 0): #print error rate every 10 epochs
        error = nn.evaluate(testData)
        print("epoch %d: %f"%(i,error[0]))
    nn.minibatch(data,10) #train model with minibatches of 10

Built With

• numpy
• python