public void TestDim1()
{
var m = new Multinomial(2018);
var cdf = new[] { 1f };
Assert.AreEqual(0, m.Sample(cdf));
Assert.AreEqual(0, m.Sample(cdf));
Assert.AreEqual(0, m.Sample(cdf));
}
public void TestDim3Unscaled()
{
var m = new Multinomial(2018);
var cdf = new[] { 0.05f, 0.15f, 0.5f };
Assert.AreEqual(2, m.Sample(cdf));
Assert.AreEqual(2, m.Sample(cdf));
Assert.AreEqual(2, m.Sample(cdf));
Assert.AreEqual(1, m.Sample(cdf));
}
public void TestDim3()
{
var m = new Multinomial(2018);
var cdf = new[] { 0.1f, 0.3f, 1.0f };
Assert.AreEqual(2, m.Sample(cdf));
Assert.AreEqual(2, m.Sample(cdf));
Assert.AreEqual(2, m.Sample(cdf));
Assert.AreEqual(1, m.Sample(cdf));
}
public void Apply(TensorProxy tensorProxy, IList <int> actionIds, Dictionary <int, ActionBuffers> lastActions)
{
var agentIndex = 0;
for (var i = 0; i < actionIds.Count; i++)
{
var agentId = actionIds[i];
if (lastActions.ContainsKey(agentId))
{
var actionBuffer = lastActions[agentId];
if (actionBuffer.IsEmpty())
{
actionBuffer = new ActionBuffers(m_ActionSpec);
lastActions[agentId] = actionBuffer;
}
var discreteBuffer = actionBuffer.DiscreteActions;
for (var j = 0; j < m_ActionSize.Length; j++)
{
ComputeCdf(tensorProxy, agentIndex, m_StartActionIndices[j], m_ActionSize[j]);
discreteBuffer[j] = m_Multinomial.Sample(m_CdfBuffer, m_ActionSize[j]);
}
}
agentIndex++;
}
}
/// <summary>
/// Draw samples from a multinomial distribution based on log-probabilities specified
/// in tensor src. The samples will be saved in the dst tensor.
/// </summary>
/// <param name="src">2-D tensor with shape batch_size x num_classes</param>
/// <param name="dst">Allocated tensor with size batch_size x num_samples</param>
/// <param name="multinomial">Multinomial object used to sample values</param>
/// <exception cref="NotImplementedException">
/// Multinomial doesn't support integer tensors
/// </exception>
/// <exception cref="ArgumentException">Issue with tensor shape or type</exception>
/// <exception cref="ArgumentNullException">
/// At least one of the tensors is not allocated
/// </exception>
public static void Eval(TensorProxy src, TensorProxy dst, Multinomial multinomial)
{
if (src.DataType != typeof(float))
{
throw new NotImplementedException("Only float tensors are currently supported");
}
if (src.valueType != dst.valueType)
{
throw new ArgumentException(
"Source and destination tensors have different types!");
}
if (src.data == null || dst.data == null)
{
throw new ArgumentNullException();
}
if (src.data.batch != dst.data.batch)
{
throw new ArgumentException("Batch size for input and output data is different!");
}
var cdf = new float[src.data.channels];
for (var batch = 0; batch < src.data.batch; ++batch)
{
// Find the class maximum
var maxProb = float.NegativeInfinity;
for (var cls = 0; cls < src.data.channels; ++cls)
{
maxProb = Mathf.Max(src.data[batch, cls], maxProb);
}
// Sum the log probabilities and compute CDF
var sumProb = 0.0f;
for (var cls = 0; cls < src.data.channels; ++cls)
{
sumProb += Mathf.Exp(src.data[batch, cls] - maxProb);
cdf[cls] = sumProb;
}
// Generate the samples
for (var sample = 0; sample < dst.data.channels; ++sample)
{
dst.data[batch, sample] = multinomial.Sample(cdf);
}
}
}
public static Func <float> CreateMultiRangeUniformSampler(IList <float> intervals, int seed)
{
//RNG
Random distr = new Random(seed);
// Will be used to normalize intervalFuncs
float sumIntervalSizes = 0;
//The number of intervals
int numIntervals = (intervals.Count / 2);
// List that will store interval lengths
float[] intervalSizes = new float[numIntervals];
// List that will store uniform distributions
IList <Func <float> > intervalFuncs = new Func <float> [numIntervals];
// Collect all intervals and store as uniform distrus
// Collect all interval sizes
for (int i = 0; i < numIntervals; i++)
{
var min = intervals[2 * i];
var max = intervals[2 * i + 1];
var intervalSize = max - min;
sumIntervalSizes += intervalSize;
intervalSizes[i] = intervalSize;
intervalFuncs[i] = () => min + (float)distr.NextDouble() * intervalSize;
}
// Normalize interval lengths
for (int i = 0; i < numIntervals; i++)
{
intervalSizes[i] = intervalSizes[i] / sumIntervalSizes;
}
// Build cmf for intervals
for (int i = 1; i < numIntervals; i++)
{
intervalSizes[i] += intervalSizes[i - 1];
}
Multinomial intervalDistr = new Multinomial(seed + 1);
float MultiRange()
{
int sampledInterval = intervalDistr.Sample(intervalSizes);
return(intervalFuncs[sampledInterval].Invoke());
}
return(MultiRange);
}
public void CanSample()
{
var n = new Multinomial(largeP, 4);
var d = n.Sample();
}
public void CanSample()
{
var n = new Multinomial(_largeP, 4);
n.Sample();
}
public void FailSampleStatic()
{
Assert.That(() => Multinomial.Sample(new Random(0), _badP, 4), Throws.ArgumentException);
}
public void CanSampleStatic()
{
Multinomial.Sample(new Random(0), _largeP, 4);
}
public void FailSampleStatic()
{
Assert.Throws <ArgumentOutOfRangeException>(() => Multinomial.Sample(new Random(0), _badP, 4));
}
public void FailSampleStatic()
{
var d = Multinomial.Sample(new Random(), badP, 4);
}
public void CanSampleStatic()
{
var d = Multinomial.Sample(new Random(), largeP, 4);
}
