public void I_should_be_able_to_save_and_load_a_pretrained_neural_network(
int inputCount,
int hiddenCount,
int outputCount)
{
// Arrange
var originalNetwork = new TinyNeuralNetwork(inputCount, hiddenCount, outputCount);
// Act
originalNetwork.Save(_filePath);
var loadedNetwork = TinyNeuralNetwork.Load(_filePath);
// Assert
loadedNetwork.Should().BeEquivalentTo(originalNetwork);
}
public void Given_a_xor_truth_table_neural_network_should_learn_the_xor_operation()
{
// Arrange
var input = new float[][]
{
new [] { 1f, 1f },
new [] { 1f, 0f },
new [] { 0f, 1f },
new [] { 0f, 0f },
};
var expected = new float[][]
{
new [] { 0f },
new [] { 1f },
new [] { 1f },
new [] { 0f },
};
var network = new TinyNeuralNetwork(2, 4, 1);
var learningRate = 1f;
for (int i = 0; i < 1000; i++)
{
for (int j = 0; j < input.Length; j++)
{
network.Train(input[j], expected[j], learningRate);
}
Shuffle(input, expected);
learningRate *= .99f;
}
// Act
var actualCategorical = input.Select(network.Predict);
// Assert
var actualNumeric = actualCategorical.Select(x => Math.Round(x[0]));
var expectedNumeric = expected.Select(x => x[0]);
actualNumeric.Should().BeEquivalentTo(expectedNumeric);
}
