private DoubleArray Predict(LightWeightGC gc)
{
var x = MakePredictionsWithoutOffSet(gc);
x = x.Add(daOffSets.GetSlice(gc.timeParameterStart, (gc.timeParameterStart + gc.Xvalues.Length - 1)));
return(x);
}
public DoubleArray MakePredictionsWithoutOffSet(LightWeightGC gc)
{
ExpParameters ep = WellParameters[gc.Name];
var b = gc.Xvalues.Select(x => Math.Exp(x * ep.GrowthRate));
DoubleArray da = DoubleArray.From(b);
da = da.Multiply(ep.InitPop);
return(da);
}
public GroupFitter(GrowthCurveCollection GCC)
{
var tmpHold = new List <LightWeightGC>();
HashSet <double> times = new HashSet <double>();
Random r = new Random();
foreach (GrowthCurve gc in GCC)
{
if (gc.ExpFit == null || gc.ExpFit.SuccessfulFit == false)
{
throw new Exception("Curve " + gc.DataSetName + " has not been fit with an exponential yet");
}
LightWeightGC tmp = new LightWeightGC()
{
Xvalues = gc.FittedXValues, YValues = gc.FittedYValues, Name = gc.DataSetName
};
tmpHold.Add(tmp);
NamesToCurves[gc.DataSetName] = tmp;
xl.AddRange(gc.FittedXValues);
yl.AddRange(gc.FittedYValues);
gc.FittedXValues.ToList().ForEach(x => times.Add(x));
gc.GroupFit = this;
}
//Now to make the data array
int ParameterArraySize = times.Count + 2 * tmpHold.Count;
var t2 = times.ToList();
data = tmpHold.ToArray();
t2.Sort();
Times = t2.ToArray();
//First items 1-n are the time point offsets, next are the GrowthRate,InitPop for the different guys
Parameters = new double[ParameterArraySize];
int ArrayStart = times.Count;
for (int i = 0; i < data.Length; i++)
{
if (i < Times.Length)
{
//Parameters[i] = 1e-3;
}
data[i].GrowthParameterStart = ArrayStart + (i * 2);
double v = r.NextDouble() < .5 ? -1.0:1.0;
Parameters[data[i].GrowthParameterStart] = GCC[i].ExpFit.GrowthRate; // +GCC[i].ExpFit.GrowthRate * (r.NextDouble() * .005) * v;
v = r.NextDouble() < .5 ? -1.0:1.0;
Parameters[data[i].GrowthParameterStart + 1] = GCC[i].ExpFit.InitialPopSize; // +GCC[i].ExpFit.InitialPopSize * (r.NextDouble() * .005) * v;
double minTime = data[i].Xvalues[0];
data[i].timeParameterStart = t2.IndexOf(minTime);
}
double val = ShoNS.Optimizer.GradTester.Test(new DiffFunc(GetDerivatives), Parameters);
double val3 = val + .01;
val3++;
FitModel();
}
public double CalculateRMSE(LightWeightGC gc)
{
var mse = CalculateResiduals(gc);
double MSE = mse.ElementMultiply(mse).Sum();
double DataPoints = gc.Xvalues.Length;
double NumParmeters = Parameters.Length;
double df = DataPoints - NumParmeters;
MSE = MSE / df;
double RMSE = Math.Sqrt(MSE);
return(RMSE);
}
public DoubleArray CalculateResiduals(LightWeightGC gc)
{
return(DoubleArray.From(gc.YValues).Subtract(Predict(gc)));
}
