/// <summary>
/// Create a bernoulli distribution.
/// </summary>
/// <param name="zeroes">The amount of zeroes in the distribution.</param>
/// <param name="ones">The amount of ones in the distribution.</param>
/// <returns>A correct distribution.</returns>
/// <exception cref="ArgumentException">Thrown when one of the chances is less than zero.</exception>
public static IDiscreteDistribution <int> Distribution(int zeroes, int ones)
{
// Needs to be more or equal to zero.
if (zeroes < 0 || ones < 0)
{
throw new ArgumentException();
}
// Empty distribution if both have no chance.
if (zeroes == 0 && ones == 0)
{
return(Empty <int> .Distribution());
}
// Singleton if only one has any chance.
if (zeroes == 0)
{
return(Singleton <int> .Distribution(1));
}
if (ones == 0)
{
return(Singleton <int> .Distribution(0));
}
return(new Bernoulli(zeroes, ones));
}
public static IDiscreteDistribution <int> Distribution(int minimum, int maximum)
{
if (minimum > maximum)
{
return(Empty <int> .Distribution());
}
if (minimum == maximum)
{
return(Singleton <int> .Distribution(minimum));
}
return(new SDU(minimum, maximum));
}
public static IDiscreteDistribution <R> Distribution(IDiscreteDistribution <A> underlying, Func <A, R> projection)
{
var result = new Projected <A, R>(underlying, projection);
if (result._weights.Count == 0)
{
return(Empty <R> .Distribution());
}
var support = result.Support().ToList();
if (support.Count == 1)
{
return(Singleton <R> .Distribution(support.First()));
}
return(result);
}
public static IDiscreteDistribution <int> Distribution(IEnumerable <int> weightCollection)
{
var weights = MathHelper.ShrinkValues(weightCollection).ToList();
// Ensure we have valid weights.
if (weights.Any(x => x < 0))
{
throw new ArgumentException();
}
// If no weights above 0, empty distribution.
if (!weights.Any(x => x > 0))
{
return(Empty <int> .Distribution());
}
weights = weights.DropEndWhile(weight => weight == 0).ToList();
if (weights.Count == 1)
{
return(Singleton <int> .Distribution(0));
}
if (weights.GetIndexIfSingle(i => i != 0, out int index))
{
return(Singleton <int> .Distribution(index));
}
if (weights.Count == 2)
{
int zeroes = weights[0];
int ones = weights[1];
return(Bernoulli.Distribution(zeroes, ones));
}
return(new WeightedInteger(weights));
}
public static IWeightedDistribution <T> Distribution(IDictionary <T, float> weights)
{
var weightArray = weights.ToArray();
switch (weightArray.Length)
{
case 0:
return(Empty <T> .Distribution());
case 1:
var item = weights.Keys.First();
return(Singleton <T> .Distribution(item));
case 2:
var item1 = weights.First();
var item2 = weights.Last();
var probability = item1.Value / (item1.Value + item2.Value);
return(Flip <T> .Distribution(item1.Key, item2.Key, probability));
default:
return(new FloatWeighted <T>(weights));
}
}
public Markov <T> ToDistribution()
{
var initial = _initial.ToDistribution();
var transitions = _transitions.ToDictionary(
kv => kv.Key,
kv => kv.Value.ToDistribution());
IDistribution <T> Transition(T state) => transitions.GetValueOrDefault(state, Empty <T> .Distribution());
return(Markov <T> .Distribution(initial, Transition));
}