XOR Problem

How Neural Network effects the ability to solve the XOR problem?

Dataset

Dataset generated to imitate the XOR problem with some noise. Can specify the number of vectors to generate.

Here is the example of using the dataset.py file:

python3 dataset.py <noise_value> <number_of_vectors>

Models

The models used are artificial neural networks. With the following specifications:

• Input nodes: 2 (x and y)
• Output nodes: 1 (0 or 1)
• Number of hidden layers: 2, 4, or 8
• Activation function: Sigmoid
• Loss function: Mean Squared Error
• Optimiser: Stochastic Gradient Descent

Results

• Running 10 tests for each model with 1000 epochs of training
• Each with a different number of training vectors