Ali AlKahki posted on 02-24-2011 4:50 AM
I am trying to implement a simple full Bayesian CRF as follwoing
double[] obsImage = new double[] { 0, 0, 1,0, 0, 1, 0, 1, 0, 1 };
bool[] trueImage = new bool[] { true, true, false, true, true, false, true, false, true, false };
int numPixels = 10;
Variable<double> probXEqualX = Variable.Beta(1, 1);
var y = new Variable<Vector>[numPixels];
var x = new Variable<bool>[numPixels];
Variable<Vector> w = Variable.Random(new VectorGaussian(Vector.Zero(2),PositiveDefiniteMatrix.Identity(2)));
for (int i = 0; i < numPixels; i++)
{
x[i] = Variable.New<bool>();
y[i] = Vector.FromArray(new double[] { 1, obsImage[i] });
double noise = .1;
Variable.ConstrainEqual(x[i]==((Variable.InnerProduct(w, y[i])) > 0),x[i]==(Variable.GaussianFromMeanAndVariance(Variable.InnerProduct(w, y[i]), noise) > 0));//,
Variable.Bernoulli(probYEqualX));
}
for (int i = 0; i < numPixels - 1; i++)
{
{
Variable.ConstrainEqual(x[i] == x[i + 1], Variable.Bernoulli(probXEqualX));
}
}
InferenceEngine engine = new InferenceEngine();
for (int i = 0; i < numPixels; i++)
{
x[i].ObservedValue = trueImage[i];
}
Console.WriteLine(engine.Infer<VectorGaussian>(w));
Console.WriteLine(engine.Infer<Beta>(probXEqualX));
which gives me :
Compiling model...done.
Iterating:
.........|.........|.........|.........|.........| 50
VectorGaussian(0 0, 1 0)
0 1
Compiling model...done.
Iterating:
.........|.........|.........|.........|.........| 50
Beta(3,8)[mean=0.2727]
Press any key to continue . . .
I can not figure out what is wrong in this code. Any help please ?
thank you
