private void GenerateStats()
{
StatsData tmpStats;
for (int i = 0; i < EM.clusters; i++)
{
tmpStats = new StatsData();
tmpStats.Variance = MyUtil.VARIANCE;
tmpStats.Mean = (float)(MyUtil.Rand.Next((int)MyUtil.MIN_MEAN, (int)MyUtil.MAX_MEAN));
tmpStats.ChangeColor();
tmpStats.GaussianPoints = Gauss.FindPoints(tmpStats.Mean, tmpStats.StdDev, MyUtil.DX, MyUtil.X_MAX, MyUtil.X_MIN);
tmpStats.Index = i;
tmpStats.Name = " EM_" + i;
stats.Add(tmpStats);
}
}
public void UpdateGaussianParameters()
{
double posteriorSum;
double tmpVariance;
double tmpMean;
posteriorSum = 0;
for (int stat = 0; stat < stats.Count; stat++)
{
//Update means
tmpMean = 0;
for (int i = 0; i < points.Count; i++)
{
tmpMean += (points[i][stat].Posterior * points[i].Value);
posteriorSum += points[i][stat].Posterior;
}
stats[stat].Mean = tmpMean / posteriorSum;
//Update variance
tmpVariance = 0;
for (int i = 0; i < points.Count; i++)
{
tmpVariance += points[i][stat].Posterior * ((points[i].Value - stats[stat].Mean) * (points[i].Value - stats[stat].Mean));
}
if (Math.Sqrt(tmpVariance / posteriorSum) < MyUtil.MAX_VARIANCE && Math.Sqrt(tmpVariance / posteriorSum) > MyUtil.MIN_VARIANCE)
{
stats[stat].Variance = tmpVariance / posteriorSum;
}
stats[stat].GaussianPoints = Gauss.FindPoints(stats[stat].Mean, stats[stat].StdDev, MyUtil.DX, MyUtil.X_MAX, MyUtil.X_MIN);
//Update Posteriors
for (int i = 0; i < points.Count; i++)
{
points[i][stat].Prior = posteriorSum / points.Count;
}
}
}
public void ComputeGaussianDensity(int index, double mean, double stdDev)
{
bayes[index].Density = Gauss.Function(value, mean, stdDev);
}