public void periodicFunction_findsClosestMinimum()
{
GaussSiedler sut = gaussSiedler((x) => Math.Sin(x[0]));
sut.MinFunctionChange = 1e-6;
sut.MinPositionChange = 1e-6;
List <double[]> initialPoints = new List <double[]>()
{
new double[] { 0.0 },
new double[] { -Math.PI * 0.5 },
new double[] { 4.0 },
new double[] { 10.0 }
};
List <double[]> expectedPoints = new List <double[]>()
{
new double[] { -Math.PI * 0.5 },
new double[] { -Math.PI * 0.5 },
new double[] { Math.PI * 1.5 }, // It fails here, as there is maximum which hides minumum, so function goes to 2nd closest minumum
new double[] { Math.PI * 3.5 }
};
for (int i = 0; i < initialPoints.Count; ++i)
{
sut.InitialPoint = new DenseVector(initialPoints[i]);
var result = sut.Solve();
Assert.AreEqual(-1.0, result.CurrentCost, error);
Assert.AreEqual(expectedPoints[i][0], result.CurrentPoint[0], error);
}
}
public void multipleInputs_multipleMinima()
{
GaussSiedler sut = gaussSiedler((x) => Math.Sin(x[0]) + Math.Cos(x[1]), 2);
List <double[]> initialPoints = new List <double[]>()
{
new double[] { -Math.PI * 0.5, Math.PI },
new double[] { -Math.PI * 0.5 + 1.0, Math.PI - 1.0 },
new double[] { Math.PI * 1.5 - 0.8, -Math.PI * 5.0 + 0.4 }
};
List <double[]> expectedPoints = new List <double[]>()
{
new double[] { -Math.PI * 0.5, Math.PI }, // x = 2kPi + 3Pi/2, y = 1kPi + Pi
new double[] { -Math.PI * 0.5, Math.PI },
new double[] { Math.PI * 1.5, -Math.PI * 5.0 }
};
for (int i = 0; i < initialPoints.Count; ++i)
{
sut.InitialPoint = new DenseVector(initialPoints[i]);
var result = sut.Solve();
Assert.AreEqual(-2.0, result.CurrentCost, error);
Assert.AreEqual(expectedPoints[i][0], result.CurrentPoint[0], error);
Assert.AreEqual(expectedPoints[i][1], result.CurrentPoint[1], error);
}
}
public void startFromMiddleOfTwoMinimas()
{
GaussSiedler sut = gaussSiedler((x) => Math.Cos(x[0]));
List <double[]> initialPoints = new List <double[]>()
{
new double[] { 0.0 },
new double[] { Math.PI * 2.0 },
};
foreach (var x0 in initialPoints)
{
sut.InitialPoint = new DenseVector(x0);
var result = sut.Solve();
Assert.AreEqual(-1.0, result.CurrentCost, error);
}
}
public void quadric()
{
GaussSiedler sut = gaussSiedler((x) => 2.0 * x[0] * x[0] - 4 * x[0] + 4);
List <double[]> initialPoints = new List <double[]>()
{
new double[] { 0.0 },
new double[] { 1.0 },
new double[] { 100.0 },
new double[] { -100.0 }
};
foreach (var x0 in initialPoints)
{
sut.InitialPoint = new DenseVector(x0);
var result = sut.Solve();
Assert.AreEqual(2.0, result.CurrentCost, error);
Assert.AreEqual(1.0, result.CurrentPoint[0], error);
}
}
public void multipleInputs_singleMinimum()
{
// f(x,y) = (x - 1) ^ 2 + (y + 2) ^ 2 + 4
GaussSiedler sut = gaussSiedler((x) => (x[0] - 1.0) * (x[0] - 1.0) + (x[1] + 2.0) * (x[1] + 2.0) + 4.0, 2);
List <double[]> initialPoints = new List <double[]>()
{
new double[] { 1.0, 2.0 },
new double[] { 0.0, 0.0 },
new double[] { 1.0, 4.0 },
new double[] { -5.0, -2.0 },
new double[] { -100.0, 100.0 },
};
for (int i = 0; i < initialPoints.Count; ++i)
{
sut.InitialPoint = new DenseVector(initialPoints[i]);
var result = sut.Solve();
Assert.AreEqual(4.0, result.CurrentCost, error);
Assert.AreEqual(1.0, result.CurrentPoint[0], error);
Assert.AreEqual(-2.0, result.CurrentPoint[1], error);
}
}
public void noMinima()
{
GaussSiedler sut = gaussSiedler((x) => x[0] * x[0] * x[0]);
// TODO: what to do in such case?
}
