public TransportSSA(ModelInfo modelInfo, float duration, int repeats, int samples)
: base(modelInfo, duration, repeats, samples)
{
Configuration config = Configuration.CurrentConfiguration;
// Transport events
diffusionReactions = model.Reactions.Where(r => !r.IsLocal).ToList();
// Reaction events
reactionReactions = model.Reactions.Where(r => r.IsLocal).ToList();
currentReactionRates = new float[reactionReactions.Count];
kernels = new float[model.Species.Count][]; // jagged array (i.e. array of arrays)
speciesIndices = new int[model.Species.Count][]; // nearest neighbors of the current species
lambda = new int[model.Species.Count]; // X(t+tau) = X(t) + lambda, diffusion changes
diffusionRates = new float[diffusionReactions.Count];
_partiallyExecutedReactionsEvents = new List <int>();
// Jagged array initialization
for (int iSpecies = 0; iSpecies < model.Species.Count; ++iSpecies)
{
int numberOfNieghbors = NumberOfNeighbors(iSpecies) + 1; // +1 for the current node + number of neighbors
kernels[iSpecies] = new float[numberOfNieghbors];
speciesIndices[iSpecies] = new int[numberOfNieghbors];
}
epsilon = config.GetParameterWithDefault("tssa.epsilon", 0.01f);
greensFunctionIterations = config.GetParameterWithDefault("tssa.greensFunctionIterations", 100);
verbose = config.GetParameterWithDefault("tssa.verbose", false);
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
// Get the transport rates
FindDiffusionRates(ref diffusionRates);
}
// ReSharper restore InconsistentNaming
// The algorithm presented here (specifically in regards to the critical
// versus noncritical reactions to avoid negative species) can be found in
// 'Avoiding Negative Populations in Explicit Poisson Tau-Leaping'
// Cao, Gillespie, Petzold. Journal of Chemical Physics.
public TauLeaping(ModelInfo modelInfo, float duration, int repeats, int samples, ModelBuilder modelBuilder = null)
: base(modelInfo, duration, repeats, samples, modelBuilder)
{
Configuration config = Configuration.CurrentConfiguration;
_currentRates = new float[model.Reactions.Count];
_jacobianMatrix = new float[model.Reactions.Count, model.Species.Count];
_nonCriticalReactions = new List <Reaction>();
_criticalReactions = new List <Reaction>();
// We'll need at most reactions.Count values in one or the other array
_criticalRates = new float[model.Reactions.Count];
_noncriticalRates = new float[model.Reactions.Count];
epsilon = config.GetParameterWithDefault("tau-leaping.epsilon", 0.001f);
nc = config.GetParameterWithDefault("tau-leaping.Nc", 2);
multiple = config.GetParameterWithDefault("tau-leaping.Multiple", 10);
SSAruns = config.GetParameterWithDefault("tau-leaping.SSARuns", 100);
CheckParameters();
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
Console.WriteLine("epsilon = {0}", epsilon);
Console.WriteLine("nc = {0}", nc);
Console.WriteLine("multiple = {0}", multiple);
Console.WriteLine("SSA runs = {0}", SSAruns);
}
public ExitTimes(ModelInfo modelInfo, double duration, int repeats, int samples)
: base(modelInfo, duration, repeats, samples, new ModelBuilder())
{
currentRates = new double[model.Reactions.Count];
lambda = new List <double>();
lambdaLists = new List <List <double> >();
exitTimes = new List <double>();
numberOfGammaRNs = 0.0;
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
Configuration config = Configuration.CurrentConfiguration;
_epsilon = Math.Min(config.GetParameterWithDefault("et.epsilon", 0.5), 1.0);
_eventName = config.GetParameterWithDefault("et.eventName", "exitTimeEvent");
_verbose = config.GetParameterWithDefault("et.verbose", false);
_testing = config.GetParameterWithDefault("et.testing", false);
_convergenceTest = config.GetParameterWithDefault("et.convergence", false);
_efficiencyCutoff = config.GetParameterWithDefault("et.efficiencyCutoff", 0.2);
targetCondition = model.Predicates.First(s => s.Name == _eventName);
sampleNumber = 1;
eventAchieved = false;
if (_verbose)
{
Console.WriteLine("\n predicate: {0} with initial evaluation: {1} \n", targetCondition.Name, targetCondition.Value);
Console.WriteLine("\n epsilon = {0} \n", _epsilon);
}
}
public BLeaping(ModelInfo modelInfo, double duration, int repeats, int samples)
: base(modelInfo, duration, repeats, samples)
{
_currentRates = new double[model.Reactions.Count];
_tau = Configuration.CurrentConfiguration.GetParameterWithDefault("b-leaping.Tau", 0.1);
_step = 1;
_deltaTau = 0.0;
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
}
public void TestGenerateExponential()
{
float[] distribution = LoadDistributionProfile <float>("distlib\\distributions\\GeneratedExponential.txt");
DistributionSampler sampler = GetAesCounterSampler();
int[] bins = SampleContinuous(() => sampler.GenerateExponential(0.5f), distribution.Length);
bool passed = ValidateDistribution(distribution, bins, 0.025f);
Assert.IsTrue(passed);
}
public void TestBinomial80Pt5()
{
float[] distribution = LoadDistributionProfile <float>("distlib\\distributions\\Binomial80Pt5.txt");
DistributionSampler sampler = GetAesCounterSampler();
int[] bins = SampleDiscrete(() => sampler.GenerateBinomial(80, 0.5f), distribution.Length);
bool passed = ValidateDistribution(distribution, bins, 0.01f, 24);
Assert.IsTrue(passed);
}
public void TestStandardExponential()
{
float[] distribution = LoadDistributionProfile <float>("distlib\\distributions\\StandardExponential.txt");
DistributionSampler sampler = GetAesCounterSampler();
int[] bins = SampleContinuous(sampler.StandardExponential, distribution.Length);
bool passed = ValidateDistribution(distribution, bins, 0.01f);
Assert.IsTrue(passed);
}
public void TestGeneratedGamma3Pt5()
{
float[] distribution = LoadDistributionProfile <float>("distlib\\distributions\\GeneratedGamma3Pt5.txt");
int binCount = distribution.Length;
DistributionSampler sampler = GetAesCounterSampler();
int[] bins = SampleContinuous(() => sampler.GenerateGamma(3.0f, 0.5f), binCount);
bool passed = ValidateDistribution(distribution, bins, 0.01f, 1);
Assert.IsTrue(passed);
}
public void TestStandardGamma14()
{
float[] distribution = LoadDistributionProfile <float>("distlib\\distributions\\StandardGammaShape14.txt");
int binCount = distribution.Length;
DistributionSampler sampler = GetAesCounterSampler();
int[] bins = SampleContinuous(() => sampler.StandardGamma(14.0f), binCount);
bool passed = ValidateDistribution(distribution, bins, 0.015f, 39);
Assert.IsTrue(passed);
}
public Normal(float mean, float variance)
{
if (variance <= 0.0f)
{
throw new ArgumentException("Variance must be greater than 0.");
}
_mean = mean;
_variance = variance;
_distributionSampler = RandLibSampler.CreateRandLibSampler(RNGFactory.GetRNG());
}
public void TestUniformClosedOpen()
{
DistributionSampler sampler = GetRandLibSampler();
bool zero;
bool one;
float sample;
bool inBounds = TestUniformSampler(sampler.GenerateUniformCO, 0.0f, 1.0f, out zero, out one, out sample);
Assert.IsTrue(inBounds, string.Format("RandLibSampler.UniformCO() failed with value {0}.", sample));
Assert.IsTrue(zero, "RandLibSampler.UniformCO() failed to generate 0.0");
Assert.IsFalse(one, "RandLibSampler.UniformCO() generated 1.0");
}
public void UniformClosedClosedMinMax()
{
DistributionSampler sampler = GetRandLibSampler();
bool min;
bool max;
float sample;
bool inBounds = TestUniformSampler(() => sampler.GenerateUniformCC(UniformMin, UniformMax), UniformMin, UniformMax, out min, out max, out sample);
Assert.IsTrue(inBounds, string.Format("RandLibSampler.UniformCC(min,max) failed with value {0}.", sample));
Assert.IsTrue(min, string.Format("RandLibSampler.UniformCC(min,max) failed to generate {0}", min));
Assert.IsTrue(max, string.Format("RandLibSampler.UniformCC(min,max) failed to generate {0}", max));
}
public void TestMultinomial4From4()
{
float[] distribution = LoadDistributionProfile <float>("distlib\\distributions\\Multinomial4From4.txt");
DistributionSampler sampler = GetAesCounterSampler();
const int nEvents = 4;
const int nCategories = 4;
var bins = SampleMultinomial(nCategories, nEvents, distribution.Length, sampler);
bool passed = ValidateDistribution(distribution, bins, 0.01f);
Assert.IsTrue(passed);
}
public RLeaping(ModelInfo modelInfo, float duration, int repeats, int samples)
: base(modelInfo, duration, repeats, samples)
{
Configuration config = Configuration.CurrentConfiguration;
reactions = new List <Reaction>(model.Reactions);
currentRates = new float[reactions.Count];
jacobianMatrix = new float[reactions.Count, model.Species.Count];
_probabilityVector = new float[reactions.Count];
_executionsPerReaction = new int[reactions.Count];
_epsilon = config.GetParameterWithDefault("r-leaping.epsilon", 0.1f);
_theta = config.GetParameterWithDefault("r-leaping.theta", 0.0f);
_sortingInterval = config.GetParameterWithDefault("r-leaping.sorting interval", 365.0f);
_verbose = config.GetParameterWithDefault("r-leaping.verbose", false);
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
}
public MidPoint(ModelInfo modelInfo, float duration, int repeats, int samples)
: base(modelInfo, duration, repeats, samples, new MpModelBuilder())
{
Configuration config = Configuration.CurrentConfiguration;
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
epsilon = config.GetParameterWithDefault("midpoint.epsilon", 0.01f);
nc = config.GetParameterWithDefault("midpoint.Nc", 2);
multiple = config.GetParameterWithDefault("midpoint.Multiple", 10);
SSAruns = config.GetParameterWithDefault("midpoint.SSARuns", 100);
CheckParameters();
Console.WriteLine("epsilon = {0}", epsilon);
Console.WriteLine("nc = {0}", nc);
Console.WriteLine("multiple = {0}", multiple);
Console.WriteLine("SSA runs = {0}", SSAruns);
}
public void TestGenerateNormal()
{
float[] distribution = LoadDistributionProfile <float>("distlib\\distributions\\GeneratedNormal.txt");
int binCount = distribution.Length;
DistributionSampler sampler = GetAesCounterSampler();
var bins = new int[binCount];
for (int i = 0; i < SampleCount; i++)
{
float sample = sampler.GenerateNormal(-2.0f, 0.5f);
int index = (int)(Math.Floor(sample * 10.0f)) + 40;
if ((index >= 0) && (index < binCount))
{
bins[index]++;
}
}
bool passed = ValidateDistribution(distribution, bins, 0.025f);
Assert.IsTrue(passed);
}
public RLeapingFast(ModelInfo modelInfo, double duration, int repeats, int samples)
: base(modelInfo, duration, repeats, samples)
{
Configuration config = Configuration.CurrentConfiguration;
reactions = new List <Reaction>(model.Reactions);
currentRates = new double[reactions.Count];
_highestOrderReaction = new int[model.Species.Count];
_muHat = new double[model.Species.Count];
_sigmaHat2 = new double[model.Species.Count];
_varHat = new double[model.Species.Count];
_probabilityVector = new double[reactions.Count];
_executionsPerReaction = new int[reactions.Count];
_epsilon = config.GetParameterWithDefault("r-leaping.epsilon", 0.1);
_sortingInterval = config.GetParameterWithDefault("r-leaping.sorting interval", 365.0);
_verbose = config.GetParameterWithDefault("r-leaping.verbose", false);
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
}
public void TestGeneratePoisson10()
{
int[] distribution = LoadDistributionProfile <int>("distlib\\distributions\\PoissonLambda10.txt");
int binCount = distribution.Length;
DistributionSampler sampler = GetAesCounterSampler();
int[] bins = SampleDiscrete(() => sampler.GeneratePoisson(10.0f), binCount);
bool passed = true;
for (int i = 0; i < binCount; i++)
{
float error = Math.Abs(distribution[i] - bins[i]) / (float)SampleCount;
if (error > 0.002502352f)
{
passed = false;
}
Console.WriteLine("{0},{1}", distribution[i], bins[i]);
}
Assert.IsTrue(passed);
}
public FractionalDiffusion(ModelInfo modelInfo, float duration, int repeats, int samples)
: base(modelInfo, duration, repeats, samples, new ModelBuilder())
{
Configuration config = Configuration.CurrentConfiguration;
#region initializeFields
_reactions = new List <Reaction>(model.Reactions);
_alpha = config.GetParameterWithDefault("fd.alpha", 0.5f);
_verbose = config.GetParameterWithDefault("fd.verbose", false);
_Dalpha = config.GetParameterWithDefault("fd.Dalpha", 1.0f);
_constant = config.GetParameterWithDefault("fd.constant", 0.25f);
_h = config.GetParameterWithDefault("fd.h", 1.0f);
_truncationDistance = config.GetParameterWithDefault("fd.truncation", (int)Math.Round(((double)(modelInfo.Locales.Count() - 1)) / 4.0));
_fourierNumber = 0.5;
_n = modelInfo.Locales.Count() - 1;
_currentRates = new float[_reactions.Count];
_field = new int[_n];
_numberOfNonDiscretizedSpecies = modelInfo.Species.Count() / _n;
_tau = (float)TimeStep(_fourierNumber, _constant, _h, _alpha, _Dalpha);
_dispersalChanges = new int[model.Species.Count];
_executionsPerReaction = new int[model.Reactions.Count];
_distributionSampler = RandLibSampler.CreateRandLibSampler(rng);
#endregion
if (_verbose == true)
{
Console.WriteLine("number of locales = {0}", modelInfo.Locales.Count() - 1);
Console.WriteLine("number of species = {0}", _numberOfNonDiscretizedSpecies);
Console.WriteLine("truncation = {0}", _truncationDistance);
Console.WriteLine("alpha = {0}", _alpha);
Console.WriteLine("D_alpha = {0}", _Dalpha);
Console.WriteLine("Fouier number (1d) = {0}", _fourierNumber);
Console.WriteLine("h = {0}", _h);
Console.WriteLine("C = {0}", _constant);
Console.WriteLine("fractional diffusion tau = {0}", _tau);
}
}
public SampleSpace4(string unused = null)
{
this._space = Space4.ZeroOne;
this._distribution_sampler = new DistributionSampler();
}
public void RandLibToStringTest()
{
DistributionSampler sampler = GetRandLibSampler();
Assert.AreEqual("RandLib Distribution Sampler (RandLib Random Variate Generator (RandLib PRNG))", sampler.ToString());
}
public void AesCounterToStringTest()
{
DistributionSampler sampler = GetAesCounterSampler();
Assert.AreEqual("RandLib Distribution Sampler (AesCounter Random Variate Generator (AesCounter PRNG))", sampler.ToString());
}
private static int[] SampleMultinomial(int nCategories, int nEvents, int binCount, DistributionSampler sampler)
{
int nTotal = nCategories * (nCategories + 1) / 2;
var probabilityVector = new float[nCategories];
for (int i = 0; i < nCategories; i++)
{
probabilityVector[i] = (float)(i + 1) / nTotal;
}
var bins = new int[binCount];
var events = new int[nCategories];
for (int i = 0; i < SampleCount; i++)
{
sampler.GenerateMultinomial(nEvents, probabilityVector, events);
int index = 0;
for (int j = nCategories - 1; j >= 0; j--)
{
index *= (nEvents + 1);
index += events[j];
}
bins[index]++;
}
return(bins);
}
private DistributionSampler GetAesCounterSampler()
{
DistributionSampler sampler = RandLibSampler.CreateRandLibSampler(AesCounterVariateGenerator.CreateAesCounterVariateGenerator(new uint[] { 20130110, 19680201 }));
return(sampler);
}
