/// <summary>
/// Compute the output
/// </summary>
/// <param name="X">input data of size (n_x,m)</param>
/// <param name="parameters">output of initialization function, [W1,b1,W2,b2]</param>
/// <returns>A2(Matrix) -- The sigmoid output of the second activation cache -- a dictionary containing "Z1", "A1", "Z2" and "A2"</returns>
public static object[] Forward_propagation(Matrix X, Dictionary <string, Matrix> parameters)
{
// Retrieve each parameter from the dictionary "parameters"
Matrix W1 = parameters["W1"];
Matrix b1 = parameters["b1"];
Matrix W2 = parameters["W2"];
Matrix b2 = parameters["b2"];
// Implement Forward Propagation to calculate A2 (probabilities)
Matrix Z1 = W1 * X + b1;
Matrix A1 = Matrix.tanh(Z1); // tanh
Matrix Z2 = W2 * A1 + b2;
Matrix A2 = LogisticRegression.Sigmoid(Z2);
Dictionary <string, Matrix> cache = new Dictionary <string, Matrix>();
cache.Add("Z1", Z1);
cache.Add("A1", A1);
cache.Add("Z2", Z2);
cache.Add("A2", A2);
object[] results = { A2, cache };
return(results);
}
public void SigmoidReturnsOpoint5ForZeroValue()
{
// arrange
// act
var result = LogisticRegression.Sigmoid(0);
// assert
Assert.AreEqual(0.5, result, 1E-13);
}
public void SigmoidReturnsCorrectResultForTestsData(double value, double expectedResult)
{
// arrange
// act
var result = LogisticRegression.Sigmoid(value);
// assert
var error = Math.Abs(expectedResult - result);
Assert.Less(error, double.Epsilon);
}
public void HypothesisReturnsCorrectResultForTestData()
{
// arrange
IList <double> xs = new double[] { -1, -2 };
IList <double> thetas = new double[] { 9, 8, 7 };
var expectedResult = LogisticRegression.Sigmoid(thetas[0] + thetas[1] * xs[0] + thetas[2] * xs[1]);
// act
var result = LogisticRegression.Hypothesis(xs, thetas);
// assert
Assert.AreEqual(expectedResult, result, 1E-13);
}
public void HypothesisReturnsThetaZeroSigmoidIfXsIsEmpty()
{
// arrange
var rnd = new Random();
var randomDouble = rnd.NextDouble();
var expectedResult = LogisticRegression.Sigmoid(randomDouble);
IList <double> xs = new double[0];
IList <double> thetas = new double[] { randomDouble };
// act
var result = LogisticRegression.Hypothesis(xs, thetas);
// assert
Assert.AreEqual(expectedResult, result, 1E-13);
}