private void OnHostPick()
{
if (state != 1)
{
return;
}
// Host cannot pick the user pick, or the car position
int hc = rnd.Next(3);
while (hc == p.ObservedValue || hc == actualCarPosition)
{
hc = rnd.Next(3);
}
doAnimation("opendoor" + (hc + 1));
doAnimation("showgoatg" + (hc + 1));
h.ObservedValue = hc;
hostIsObserved.ObservedValue = true;
cProbs = getProbs(c);
string fmtStr = "{0:0.0000}";
Lbl1.Content = String.Format(fmtStr, cProbs[0]);
Lbl2.Content = String.Format(fmtStr, cProbs[1]);
Lbl3.Content = String.Format(fmtStr, cProbs[2]);
UpdateDoors(cProbs);
label1.Content = "P ( Car Inside )";
button2.Visibility = Visibility.Hidden;
state = 2;
}
private void UpdateDoors(Microsoft.ML.Probabilistic.Math.Vector cProbs)
{
for (int a = 0; a < 3; a++)
{
if (cProbs[a] == 1)
{
doAnimation("showcarg" + (a + 1));
doAnimation("opendoor" + (a + 1));
}
}
}
private void BeforeHostPick()
{
h.ClearObservedValue();
hostIsObserved.ObservedValue = false;
Microsoft.ML.Probabilistic.Math.Vector cProbs = getProbs(c);
Microsoft.ML.Probabilistic.Math.Vector hProbs = getProbs(h);
string fmtStr = "{0:0.0000}\n{1:0.0000}";
Lbl1.Content = String.Format(fmtStr, cProbs[0], hProbs[0]);
Lbl2.Content = String.Format(fmtStr, cProbs[1], hProbs[1]);
Lbl3.Content = String.Format(fmtStr, cProbs[2], hProbs[2]);
UpdateDoors(cProbs);
state = 1;
}
/// <summary>
/// 根据均值和协方差数据生成高斯对象
/// </summary>
public void InitGaussian()
{
if (gaussian != null)
{
return;
}
Microsoft.ML.Probabilistic.Math.Vector mean = null;
Microsoft.ML.Probabilistic.Math.PositiveDefiniteMatrix covar = null;
try
{
mean = this.means.toMathVector();
covar = new Microsoft.ML.Probabilistic.Math.PositiveDefiniteMatrix(covariance);
gaussian = VectorGaussian.FromMeanAndVariance(mean, covar);
checkGaussian();
}
catch (Exception e)
{
//这里异常的主要原因是得到的协方差矩阵不是正定矩阵。
//This happens if the diagonal values of the covariance matrix are (very close to) zero.
//A simple fix is add a very small constant number to c.
//logger.Error(e.Message);
for (int i = 0; i < covariance.GetLength(0); i++)
{
for (int j = 0; j < covariance.GetLength(1); j++)
{
if (i == j)
{
covariance[i, j] += 0.001;
}
//if (i == j && covariance[i, j] < 0) covariance[i, j] += 0.01;
//if (i != j) covariance[i,j] = 0;
}
}
covar = new Microsoft.ML.Probabilistic.Math.PositiveDefiniteMatrix(covariance);
gaussian = VectorGaussian.FromMeanAndVariance(mean, covar);
checkGaussian();
}
}
/// <summary>
/// 高斯分布概率值
/// </summary>
/// <param name="values"></param>
/// <returns></returns>
public double Prob(List <Vector> values)
{
this.InitGaussian();
Microsoft.ML.Probabilistic.Math.Vector v = values.toMathVector();
return(Math.Exp(gaussian.GetLogProb(v)));
}
internal bool Reclassify()
{
if (classifying)
{
return(false);
}
classifying = true;
List <Item> positive = new List <Item>();
List <Item> negative = new List <Item>();
foreach (Item item in Items)
{
if (item.State == 1)
{
positive.Add(item);
}
if (item.State == -1)
{
negative.Add(item);
}
}
int count = positive.Count + negative.Count;
if (count == 0)
{
Reset();
classifying = false;
return(false);
}
Vector[] trainData = new Vector[count];
bool[] trainLabels = new bool[count];
Vector[] testData = new Vector[Items.Count - count];
int trainCount = 0, testCount = 0;
foreach (Item item in Items)
{
if (item.State != 0)
{
trainData[trainCount] = item.Data;
trainLabels[trainCount] = item.State == 1;
trainCount++;
}
else
{
testData[testCount++] = item.Data;
}
}
form1.bpm.Train(trainData, trainLabels);
double[] probs = form1.bpm.Test(testData);
testCount = 0;
foreach (Item item in Items)
{
if (item.State == 0)
{
item.probTrue = probs[testCount++];
}
}
classifying = false;
return(true);
}
public Item(string filename, Vector data)
{
this.Filename = filename;
this.Data = data;
}
