/// <summary>
/// Optimization method. This calls MPFit Levenberg Marquardt solver with constraints.
/// </summary>
/// <param name="a">optimization parameter initial guess</param>
/// <param name="ia">accompanying array to <paramref name="a"/> that specifies which parameters to fit (held constant otherwise)</param>
/// <param name="lowerBounds">accompanying array that specifies lower bounds for parameters</param>
/// <param name="upperBounds">accompanying array that specifies upper bounds</param>
/// <param name="y">"measured" values</param>
/// <param name="ey">standard deviation values of <paramref name="y"/></param>
/// <param name="forwardFunc">delegate function that evaluates the objective function given a parameter optimization array and (optional) constant variables</param>
/// <param name="forwardVariables"></param>
public double[] SolveWithConstraints(double[] a, bool[] ia, double[] lowerBounds, double[] upperBounds, double[] y, double[] ey, Func <double[], object[], double[]> forwardFunc, params object[] forwardVariables)
{
var data = new OptimizationData
{
Y = y,
Ey = ey,
ForwardFunc = forwardFunc,
ForwardVariables = forwardVariables
};
mp_par[] pars = a.Select((ai, i) => new mp_par {
isFixed = ia[i] ? 0 : 1
}).ToArray();
for (int i = 0; i < pars.Length; i++)
{
pars[i].limited[0] = 1; // specify lower bound exists
pars[i].limited[1] = 1; // specify upper bound exists
pars[i].limits[0] = lowerBounds[i];
pars[i].limits[1] = upperBounds[i];
}
mp_result result = new mp_result(a.Length);
int status = MPFit.Solve(MPFitFunc, data.Y.Length, pars.Length, a, pars, null, data, ref result);
return(a);
}
/// <summary>
/// Standard function prototype that MPFit knows how to call. Use <paramref name="vars"/> to store information reqired to evaluate any objective function
/// </summary>
/// <param name="parameters">array of fit parameters</param>
/// <param name="dy">array of residuals to be returned</param>
/// <param name="dvec">not used</param>
/// <param name="vars">private data</param>
/// <returns></returns>
private static int MPFitFunc(double[] parameters, double[] dy, IList <double>[] dvec, object vars)
{
OptimizationData oData = vars as OptimizationData;
if (oData != null)
{
var differenceValues = oData
.ForwardFunc(parameters, oData.ForwardVariables)
.Zip(oData.Y, oData.Ey, (fi, yi, eyi) => eyi != 0 ? (yi - fi) / eyi : 0.0);
dy.PopulateFromEnumerable(differenceValues);
return(0); // success
}
return(MPFit.MP_ERR_PARAM);
}
/// <summary>
/// Optimization method. This calls MPFit Levenberg Marquardt solver.
/// For examples of usage, see TestMPFit.cs:
/// https://csmpfit.codeplex.com/SourceControl/latest#src/MPFitLib.Test/TestMPFit.cs
/// For an example of a call with the objective function:
/// https://csmpfit.codeplex.com/SourceControl/latest#src/MPFitLib.Test/ForwardModels.cs
/// </summary>
/// <param name="a">optimization parameter initial guess</param>
/// <param name="ia">accompanying array to <paramref name="a"/> that specifies which parameters to fit (held constant otherwise)</param>
/// <param name="y">"measured" values</param>
/// <param name="ey">standard deviation values of <paramref name="y"/></param>
/// <param name="forwardFunc">delegate function that evaluates the objective function given a parameter optimization array and (optional) constant variables</param>
/// <param name="forwardVariables"></param>
public double[] Solve(double[] a, bool[] ia, double[] y, double[] ey, Func <double[], object[], double[]> forwardFunc, params object[] forwardVariables)
{
var data = new OptimizationData
{
Y = y,
Ey = ey,
ForwardFunc = forwardFunc,
ForwardVariables = forwardVariables
};
mp_par[] pars = a.Select((ai, i) => new mp_par {
isFixed = ia[i] ? 0 : 1
}).ToArray();
mp_result result = new mp_result(a.Length);
int status = MPFit.Solve(MPFitFunc, data.Y.Length, pars.Length, a, pars, null, data, ref result);
return(a);
}
