/// <summary>
/// Applies this decider's feature to all of the points, storing the value in the <see cref="T:ImageDataPoint.FeatureValue"/> property.
/// </summary>
/// <param name="points">Points to use when applying the feature</param>
public void ApplyFeature(List <T> points)
{
int count = points.Count;
for (int i = 0; i < count; i++)
{
points[i].FeatureValue = _feature.Compute(points[i]);
}
}
public void ApplyFeature(List <T> points)
{
foreach (T point in points)
{
point.FeatureValue = _feature.Compute(point);
}
}
/// <summary>
/// Computes the factor potential function for the given parameters.
/// </summary>
///
/// <param name="previousState">The previous state in a given sequence of states.</param>
/// <param name="currentState">The current state in a given sequence of states.</param>
/// <param name="observations">The observation vector.</param>
/// <param name="index">The index of the observation in the current state of the sequence.</param>
/// <param name="outputClass">The output class label for the sequence.</param>
/// <returns>The value of the factor potential function evaluated for the given parameters.</returns>
///
public virtual double Compute(int previousState, int currentState,
T[] observations, int index, int outputClass = 0)
{
if (outputClass != this.Index)
{
return(Double.NegativeInfinity);
}
int start = FactorParameters.Offset;
int end = FactorParameters.Offset + FactorParameters.Count;
double sum = 0;
for (int k = start; k < end; k++)
{
double weight = Owner.Weights[k];
if (Double.IsNaN(weight))
{
Owner.Weights[k] = weight = 0;
}
if (weight != 0)
{
IFeature <T> feature = Owner.Features[k];
double value = feature.Compute(previousState, currentState,
observations, index, outputClass);
if (value != 0)
{
sum += weight * value;
}
}
if (Double.IsNaN(sum))
{
return(0);
}
if (Double.IsPositiveInfinity(sum))
{
return(Double.MaxValue);
}
if (Double.IsNegativeInfinity(sum))
{
return(Double.NegativeInfinity);
}
}
return(sum);
}
private static void fillFeatureValues(IFeatureFactory <T, float[]> factory, int numFeatures, List <T> data)
{
_numFeatures = numFeatures;
_buildFeatures = new IFeature <T, float[]> [numFeatures];
_featureValues = new float[numFeatures][];
for (int i = 0; i < numFeatures; i++)
{
IFeature <T, float[]> feature = factory.Create();
_featureValues[i] = data.Select(o => feature.Compute(o)).ToArray();
_buildFeatures[i] = feature;
}
}
/// <summary>
/// Test metoda. Potrebne su dvije ulazne datoteke datoteke, InputFile i OutputFile. U jednoj su
/// ulazni podatci za značajku, a u drugoj referentni podatci (izlaz iz Matlaba).
/// Ulazni podatci i izlaz i matlaba se spremaju koristeći sljedeću sintaksu:
/// save('ime_datoteke.txt','varijabla','-ascii','-double','-tabs');
/// </summary>
/// <param name="InputFile">Datoteka s ulaznim podatcima</param>
/// <param name="OutputFile">Datoteka s izlaznim podatcima</param>
/// <param name="DataProvider">DataProvider za IFeature</param>
/// <param name="Feature">Značajka</param>
/// <param name="SubWindowLength">Dužina prozora u uzorcima</param>
/// <param name="SubWindowShift">Dužina pomaka u uzorcima</param>
/// <returns></returns>
public static TestResult ExecuteTest(string InputFile, string OutputFile, IDataProvider DataProvider, IFeature Feature, int SubWindowLength, int SubWindowShift)
{
List <double> InputData = ReadFromFile(InputFile);
List <double> OutputData = ReadFromFile(OutputFile);
FastQueue <double> Data = new FastQueue <double>(100000);
List <double> Result = new List <double>();
Data.Enqueue(InputData.ToArray());
while (Data.Count > SubWindowLength)
{
List <double> CurrentData;
CurrentData = Data.Peek(SubWindowLength);
Data.Delete(SubWindowShift);
DataProvider.Data = CurrentData;
Feature.Compute();
Result.Add(Feature.Feature);
}
double absdiff = 0;
for (int i = 0; i < Result.Count || i < OutputData.Count; i++)
{
absdiff += Math.Abs(Result[i] - OutputData[i]);
}
TestResult tr = new TestResult();
tr.AbsoluteDifference = absdiff;
if (absdiff < 100)
{
tr.Success = true;
}
else
{
tr.Success = false;
}
return(tr);
}
/// <summary>
/// Test metoda. Potrebne su dvije ulazne datoteke datoteke, InputFile i OutputFile. U jednoj su
/// ulazni podatci za značajku, a u drugoj referentni podatci (izlaz iz Matlaba).
/// Ulazni podatci i izlaz i matlaba se spremaju koristeći sljedeću sintaksu:
/// save('ime_datoteke.txt','varijabla','-ascii','-double','-tabs');
/// </summary>
/// <param name="InputFile">Datoteka s ulaznim podatcima</param>
/// <param name="OutputFile">Datoteka s izlaznim podatcima</param>
/// <param name="DataProvider">DataProvider za IFeature</param>
/// <param name="Feature">Značajka</param>
/// <param name="SubWindowLength">Dužina prozora u uzorcima</param>
/// <param name="SubWindowShift">Dužina pomaka u uzorcima</param>
/// <returns></returns>
public static TestResult ExecuteTest(string InputFile, string OutputFile, IDataProvider DataProvider, IFeature Feature, int SubWindowLength, int SubWindowShift)
{
List<double> InputData = ReadFromFile(InputFile);
List<double> OutputData = ReadFromFile(OutputFile);
FastQueue<double> Data = new FastQueue<double>(100000);
List<double> Result = new List<double>();
Data.Enqueue(InputData.ToArray());
while (Data.Count > SubWindowLength)
{
List<double> CurrentData;
CurrentData = Data.Peek(SubWindowLength);
Data.Delete(SubWindowShift);
DataProvider.Data = CurrentData;
Feature.Compute();
Result.Add(Feature.Feature);
}
double absdiff = 0;
for (int i = 0; i < Result.Count || i < OutputData.Count; i++)
{
absdiff += Math.Abs(Result[i] - OutputData[i]);
}
TestResult tr = new TestResult();
tr.AbsoluteDifference = absdiff;
if (absdiff < 100)
tr.Success = true;
else
tr.Success = false;
return tr;
}
private static Node compute(List <T> data, int NCount, IFeatureFactory <T, float[]> factory, int numFeatures, int numThresholds, float min_rd, int min_y, int currentDepth, int maxDepth)
{
if (currentDepth == maxDepth - 1)
{
return new Node {
NodeType = NodeType.Leaf
}
}
;
int YCount = data.Count;
if (NCount < YCount)
{
NCount = YCount;
}
Split best = new Split {
Score = float.MinValue
};
for (int i = 0; i < numFeatures; i++)
{
IFeature <T, float[]> feature = factory.Create();
var featureValues = from point in data
select feature.Compute(point);
Split split = findBestSplit(featureValues.OrderBy(o => o).ToArray(), min_rd, min_y, NCount, numThresholds);
if (split.Score > best.Score)
{
best = split;
best.Feature = feature;
}
}
if (best.Feature == null)
{
return new Node {
NodeType = NodeType.Leaf
}
}
;
Node node = new Node {
NodeType = NodeType.Branch, Feature = best.Feature, Threshold = best.Threshold
};
List <T> left = new List <T>();
List <T> right = new List <T>();
foreach (T point in data)
{
if (best.Feature.Compute(point) < best.Threshold)
{
left.Add(point);
}
else
{
right.Add(point);
}
}
if (left.Count == 0 || right.Count == 0)
{
return new Node {
NodeType = NodeType.Leaf
}
}
;
UpdateManager.WriteLine("{0}:{1} {2}|{3} {4}", currentDepth, best.Score, left.Count, right.Count, best.Feature);
node.Left = compute(left, best.NLeft, factory, numFeatures, numThresholds, min_rd, min_y, currentDepth + 1, maxDepth);
node.Right = compute(right, NCount - best.NLeft, factory, numFeatures, numThresholds, min_rd, min_y, currentDepth + 1, maxDepth);
return(node);
}
}
}
private double[] gradient(T[][] observations, int[][] labels, double[] g)
{
var model = Model;
var function = model.Function;
int states = model.States;
int n = observations.Length;
int d = Model.Function.Weights.Length;
int Tmax = observations.Max(x => x.Length);
int progress = 0;
g.Clear();
// Compute sequence probabilities
Parallel.For(0, observations.Length, ParallelOptions,
() =>
{
// Create thread-local storage
var work = new double[states + 1, states][];
for (int j = 0; j < states + 1; j++)
{
for (int k = 0; k < states; k++)
{
work[j, k] = new double[Tmax];
}
}
return(new
{
bwd = new double[Tmax, states],
fwd = new double[Tmax, states],
sum1 = new double[d],
sum2 = new double[d],
work = work,
count = new int[] { 0 }
});
},
(i, state, local) =>
{
T[] x = observations[i];
var fwd = local.fwd;
var bwd = local.bwd;
var sum1 = local.sum1;
var sum2 = local.sum2;
var work = local.work;
ForwardBackwardAlgorithm.Forward(function.Factors[0], x, fwd);
ForwardBackwardAlgorithm.Backward(function.Factors[0], x, bwd);
double z = partition(fwd, x);
for (int prev = -1; prev < states; prev++)
{
for (int next = 0; next < states; next++)
{
double[] Pis = work[prev + 1, next];
for (int t = 0; t < x.Length; t++)
{
Pis[t] = p(prev, next, x, t, fwd, bwd, function) / z;
}
}
}
// Compute the gradient w.r.t. each feature
// function in the model's potential function.
int[] y = labels[i];
Parallel.For(0, g.Length, ParallelOptions, k =>
{
IFeature <T> feature = function.Features[k];
// Compute first term of the partial derivative
sum1[k] += feature.Compute(-1, y[0], x, 0);
for (int t = 1; t < x.Length; t++)
{
sum1[k] += feature.Compute(y[t - 1], y[t], x, t);
}
// Compute second term of the partial derivative
for (int prev = -1; prev < states; prev++)
{
for (int next = 0; next < states; next++)
{
double[] Pis = work[prev + 1, next];
for (int t = 0; t < Pis.Length; t++)
{
sum2[k] += feature.Compute(prev, next, x, t) * Pis[t];
}
}
}
});
local.count[0]++;
return(local);
},
(local) =>
{
lock (g)
{
for (int k = 0; k < g.Length; k++)
{
g[k] -= (local.sum1[k] - local.sum2[k]);
}
progress += local.count[0];
}
}
);
return(g);
}
private double[] gradient(int[][] observations, int[][] labels)
{
int N = observations.Length;
var function = model.Function;
var states = model.States;
double[] g = new double[function.Weights.Length];
// Compute sequence probabilities
var P = new double[N][, ][];
for (int i = 0; i < N; i++)
{
var Pi = P[i] = new double[states + 1, states][];
int[] x = observations[i];
var fwd = ForwardBackwardAlgorithm.Forward(function, x);
var bwd = ForwardBackwardAlgorithm.Backward(function, x);
double z = partition(fwd, x);
for (int prev = -1; prev < states; prev++)
{
for (int next = 0; next < states; next++)
{
double[] Pis = new double[x.Length];
for (int t = 0; t < x.Length; t++)
{
Pis[t] = p(prev, next, x, t, fwd, bwd, function) / z;
}
Pi[prev + 1, next] = Pis;
}
}
}
// Compute the gradient w.r.t. each feature
// function in the model's potential function.
for (int k = 0; k < g.Length; k++)
{
IFeature feature = function.Features[k];
double sum1 = 0.0, sum2 = 0.0;
for (int i = 0; i < N; i++)
{
int[] x = observations[i];
int[] y = labels[i];
var Pi = P[i];
// Compute first term of the partial derivative
sum1 += feature.Compute(-1, y[0], x, 0);
for (int t = 1; t < x.Length; t++)
{
sum1 += feature.Compute(y[t - 1], y[t], x, t);
}
// Compute second term of the partial derivative
for (int prev = -1; prev < states; prev++)
{
for (int next = 0; next < states; next++)
{
double[] Pis = Pi[prev + 1, next];
for (int t = 0; t < Pis.Length; t++)
{
sum2 += feature.Compute(prev, next, x, t) * Pis[t];
}
}
}
}
g[k] = -(sum1 - sum2);
}
return(g);
}
