public void Binomial(float p)
{
var weightSum = 0f;
var itemWeights = new float[10];
var n = itemWeights.Length - 1;
for (var i = 0; i < itemWeights.Length; i++)
{
itemWeights[i] = (float)Stat.BinomialDistributionMass(i, n, p);
weightSum += itemWeights[i];
}
var seed = (uint)Environment.TickCount;
var random = new SimpleRandom {
Seed = seed
};
var selectedCounts = new int[itemWeights.Length];
var selectionCount = 500000;
for (var i = 0; i < selectionCount; i++)
{
var selectedIndex = random.NextBinomial(p, n);
selectedCounts[selectedIndex]++;
}
AssertDistribution(itemWeights, weightSum, selectedCounts, selectionCount, seed);
}
public void NoStreakBool(int p)
{
var itemWeights = new[] { p, 100f - p };
var maxPositiveStreak = 2;
var maxNegativeStreak = 2;
if (p == 0)
{
maxPositiveStreak = 0;
maxNegativeStreak = int.MaxValue;
}
else if (p == 100)
{
maxPositiveStreak = int.MaxValue;
maxNegativeStreak = 0;
}
else if (p > 50)
{
maxPositiveStreak = (int)Math.Ceiling((100f / (100 - p)) - 1) * 2;
}
else
{
maxNegativeStreak = (int)Math.Ceiling((100f / p) - 1) * 2;
}
var seed = (uint)Environment.TickCount;
var random = new SimpleRandom {
Seed = seed
};
var selectedCounts = new int[itemWeights.Length];
var selectionCount = 500000;
var entropy = 0;
var positiveStreak = 0;
var negativeStreak = 0;
for (var i = 0; i < selectionCount; i++)
{
var result = random.NextBool(p, ref entropy);
if (result)
{
selectedCounts[0]++;
positiveStreak++;
Assert.True(negativeStreak <= maxNegativeStreak,
$"Found negative streak of length {negativeStreak} at {i} for seed {seed}. Max streak expected {maxNegativeStreak}.");
negativeStreak = 0;
}
else
{
selectedCounts[1]++;
negativeStreak++;
Assert.True(positiveStreak <= maxPositiveStreak,
$"Found positive streak of length {positiveStreak} at {i} for seed {seed}. Max streak expected {maxPositiveStreak}.");
positiveStreak = 0;
}
}
AssertDistribution(itemWeights, 100f, selectedCounts, selectionCount, seed);
}
public void Pick_predicate()
{
var weightSum = 0f;
var items = new float[10];
for (var i = 0; i < items.Length; i++)
{
items[i] = 1;
weightSum += items[i];
}
var seed = (uint)Environment.TickCount;
var random = new SimpleRandom {
Seed = seed
};
var selectedCounts = new int[items.Length];
var selectionCount = 10000;
for (var i = 0; i < selectionCount; i++)
{
var orderedItems = random.WeightedOrder(items, f => f, (f, index) => (f, index));
var selectedIndex = orderedItems.First(tuple => tuple.index % 2 == 0).index;
selectedCounts[selectedIndex]++;
}
var tolerance = 1.5 * items.Length / Math.Sqrt(selectionCount);
for (var i = 0; i < items.Length; i++)
{
var expected = i % 2 == 0 ? 2.0 * items[i] / weightSum : 0;
var actual = (double)selectedCounts[i] / selectionCount;
if (!(Math.Abs(expected - actual) <= tolerance * expected))
{
var builder = new StringBuilder();
builder.Append(
$"Expected {expected}, actual {actual}, tolerance {tolerance * expected} at index {i}, seed {seed}");
builder.AppendLine("\tIndex:\tfrequency,\tweight:");
for (var j = 0; j < items.Length; j++)
{
builder.AppendLine($"\t{j}:\t{selectedCounts[j]},\t{items[j]}");
}
Assert.False(true, builder.ToString());
}
}
}
private void TestPick(float[] itemWeights, float weightSum)
{
var seed = (uint)Environment.TickCount;
var random = new SimpleRandom {
Seed = seed
};
var selectedCounts = new int[itemWeights.Length];
var selectionCount = 100000;
for (var i = 0; i < selectionCount; i++)
{
var selectedIndex = random.Pick(itemWeights, f => f, (_, index) => index);
selectedCounts[selectedIndex]++;
}
AssertDistribution(itemWeights, weightSum, selectedCounts, selectionCount, seed);
}
