public WittenBellProbabilityDistribution(FrequencyDistribution <TSample> freqDist, int binCount)
{
if (binCount <= freqDist.ObservedSamples.Count)
{
throw new ArgumentOutOfRangeException("binCount");
}
_freqDist = freqDist;
if (freqDist.ObservedSamples.Count > 0)
{
int z = binCount - _freqDist.ObservedSamples.Count;
_probZero = (double)_freqDist.ObservedSamples.Count / (z * (_freqDist.SampleOutcomeCount + _freqDist.ObservedSamples.Count));
}
}
public LidstoneProbabilityDistribution(FrequencyDistribution <TSample> freqDist, double gamma, int binCount)
{
if (binCount <= freqDist.ObservedSamples.Count)
{
throw new ArgumentOutOfRangeException("binCount");
}
if (double.IsInfinity(gamma) || double.IsNaN(gamma) || gamma < 0)
{
throw new ArgumentOutOfRangeException("gamma");
}
_freqDist = freqDist;
_gamma = gamma;
_binCount = binCount;
_divisor = _freqDist.SampleOutcomeCount + (_binCount * gamma);
}
public void FixtureSetUp()
{
_fd = new FrequencyDistribution <string>();
_fd.Increment("a", 1);
_fd.Increment("b", 1);
_fd.Increment("c", 2);
_fd.Increment("d", 3);
_fd.Increment("e", 4);
_fd.Increment("f", 4);
_fd.Increment("g", 4);
_fd.Increment("h", 5);
_fd.Increment("i", 5);
_fd.Increment("j", 6);
_fd.Increment("k", 6);
_fd.Increment("l", 6);
_fd.Increment("m", 7);
_fd.Increment("n", 7);
_fd.Increment("o", 8);
_fd.Increment("p", 9);
_fd.Increment("q", 10);
}
public FrequencyDistribution(FrequencyDistribution<TSample> fd)
{
_sampleCounts = new Dictionary<TSample, int>(fd._sampleCounts);
SampleOutcomeCount = fd.SampleOutcomeCount;
}
public MaxLikelihoodProbabilityDistribution(FrequencyDistribution <TSample> freqDist)
{
_freqDist = freqDist;
}
public SimpleGoodTuringProbabilityDistribution(FrequencyDistribution <TSample> freqDist, int binCount)
{
if (binCount <= freqDist.ObservedSamples.Count)
{
throw new ArgumentOutOfRangeException("binCount");
}
_freqDist = freqDist;
_binCount = binCount;
_probs = new Dictionary <int, double>();
if (freqDist.ObservedSamples.Count == 0)
{
return;
}
var r = new List <int>();
var nr = new List <int>();
int b = 0, i = 0;
while (b != _freqDist.ObservedSamples.Count)
{
int nri = _freqDist.ObservedSamples.Count(s => _freqDist[s] == i);
if (nri > 0)
{
b += nri;
r.Add(i);
nr.Add(nri);
}
i++;
}
var zr = new double[r.Count];
var logr = new double[r.Count];
var logzr = new double[r.Count];
for (int j = 0; j < r.Count; j++)
{
i = j > 0 ? r[j - 1] : 0;
int k = j == r.Count - 1 ? 2 * r[j] - i : r[j + 1];
zr[j] = 2.0 * nr[j] / (k - i);
logr[j] = Math.Log(r[j]);
logzr[j] = Math.Log(zr[j]);
}
double xycov = 0, xvar = 0, xmean = 0, ymean = 0;
for (int j = 0; j < r.Count; j++)
{
xmean += logr[j];
ymean += logzr[j];
}
xmean /= r.Count;
ymean /= r.Count;
for (int j = 0; j < logr.Length; j++)
{
xycov += (logr[j] - xmean) * (logzr[j] - ymean);
xvar += Math.Pow(logr[j] - xmean, 2);
}
_slope = Math.Abs(xvar - 0) > double.Epsilon ? xycov / xvar : 0;
_intercept = ymean - _slope * xmean;
var rstar = new double[r.Count];
for (int j = 0; j < r.Count; j++)
{
double smoothRstar = (r[j] + 1) * GetSmoothedSamplesCount(r[j] + 1) / GetSmoothedSamplesCount(r[j]);
if (r.Count == j + 1 || r[j + 1] != r[j] + 1)
{
rstar[j] = smoothRstar;
}
else
{
double unsmoothRstar = (double)(r[j] + 1) * nr[j + 1] / nr[j];
double std = Math.Sqrt(GetVariance(r[j], nr[j], nr[j + 1]));
if (Math.Abs(unsmoothRstar - smoothRstar) <= 1.96 * std)
{
rstar[j] = smoothRstar;
}
else
{
rstar[j] = unsmoothRstar;
}
}
}
double samplesCountPrime = 0;
for (int j = 0; j < r.Count; j++)
{
samplesCountPrime += nr[j] * rstar[j];
}
_probZero = (double)_freqDist.ObservedSamples.Count(s => _freqDist[s] == 1) / _freqDist.SampleOutcomeCount;
for (int j = 0; j < r.Count; j++)
{
_probs[r[j]] = (1.0 - _probZero) * rstar[j] / samplesCountPrime;
}
}
