/// <summary>
/// Converts a variable declaration by creating definition, marginal and uses channel variables.
/// </summary>
protected override IExpression ConvertVariableDeclExpr(IVariableDeclarationExpression ivde)
{
IVariableDeclaration ivd = ivde.Variable;
VariableInformation vi = VariableInformation.GetVariableInformation(context, ivd);
// If the variable is deterministic, return
if (!vi.IsStochastic)
{
ProcessConstant(ivd);
context.OutputAttributes.Set(ivd, new DescriptionAttribute("The constant '" + ivd.Name + "'"));
return(ivde);
}
bool suppressVariableFactor = context.InputAttributes.Has <SuppressVariableFactor>(ivd);
bool isDerived = context.InputAttributes.Has <DerivedVariable>(ivd);
bool isConstant = false;
bool isInferred = context.InputAttributes.Has <IsInferred>(ivd);
int useCount;
ChannelAnalysisTransform.UsageInfo info;
if (!analysis.usageInfo.TryGetValue(ivd, out info))
{
useCount = 0;
}
else
{
useCount = info.NumberOfUsesOld;
}
if (!(algorithm is Algorithms.GibbsSampling) && !isConstant && !suppressVariableFactor && (useCount == 1) && !isInferred && isDerived)
{
// this is optional
suppressVariableFactor = true;
context.InputAttributes.Set(ivd, new SuppressVariableFactor());
}
context.InputAttributes.Remove <LoopContext>(ivd);
context.InputAttributes.Set(ivd, new LoopContext(context));
// Create variable-to-channel information for the variable.
VariableToChannelInformation vtc = new VariableToChannelInformation();
vtc.shareAllUses = (useCount == 1);
Context.InputAttributes.Set(ivd, vtc);
// Ensure the marginal prototype is set.
MarginalPrototype mpa = Context.InputAttributes.Get <MarginalPrototype>(ivd);
try
{
vi.SetMarginalPrototypeFromAttribute(mpa);
}
catch (ArgumentException ex)
{
Error(ex.Message);
}
// Create the definition channel
vtc.defChannel = ChannelInfo.DefChannel(vi);
vtc.defChannel.decl = ivd;
// Always create a variable factor for a stochastic variable
if (!isConstant && !suppressVariableFactor)
{
vi.DefineAllIndexVars(context);
IList <IStatement> stmts = Builder.StmtCollection();
// Create marginal channel
vtc.marginalChannel = ChannelInfo.MarginalChannel(vi);
vtc.marginalChannel.decl = vi.DeriveIndexedVariable(stmts, context, vi.Name + "_marginal");
context.InputAttributes.CopyObjectAttributesTo <InitialiseTo>(vi.declaration, context.OutputAttributes, vtc.marginalChannel.decl);
context.OutputAttributes.Set(vtc.marginalChannel.decl, vtc.marginalChannel);
context.OutputAttributes.Set(vtc.marginalChannel.decl, new DescriptionAttribute("marginal of '" + ivd.Name + "'"));
SetMarginalPrototype(vtc.marginalChannel.decl);
if (algorithm is GibbsSampling && ((GibbsSampling)algorithm).UseSideChannels)
{
Type marginalType = MessageTransform.GetDistributionType(vi.varType, vi.InnermostElementType,
vi.marginalPrototypeExpression.GetExpressionType(), true);
Type domainType = ivd.VariableType.DotNetType;
vtc.samplesChannel = ChannelInfo.MarginalChannel(vi);
vtc.samplesChannel.decl = vi.DeriveIndexedVariable(stmts, context, vi.Name + "_samples");
context.OutputAttributes.Remove <InitialiseTo>(vtc.samplesChannel.decl);
context.OutputAttributes.Set(vtc.samplesChannel.decl, vtc.samplesChannel);
context.OutputAttributes.Set(vtc.samplesChannel.decl, new DescriptionAttribute("samples of '" + ivd.Name + "'"));
Type samplesType = typeof(List <>).MakeGenericType(domainType);
IExpression samples_mpe = Builder.NewObject(samplesType);
VariableInformation samples_vi = VariableInformation.GetVariableInformation(context, vtc.samplesChannel.decl);
samples_vi.marginalPrototypeExpression = samples_mpe;
vtc.conditionalsChannel = ChannelInfo.MarginalChannel(vi);
vtc.conditionalsChannel.decl = vi.DeriveIndexedVariable(stmts, context, vi.Name + "_conditionals");
context.OutputAttributes.Remove <InitialiseTo>(vtc.conditionalsChannel.decl);
context.OutputAttributes.Set(vtc.conditionalsChannel.decl, vtc.conditionalsChannel);
context.OutputAttributes.Set(vtc.conditionalsChannel.decl, new DescriptionAttribute("conditionals of '" + ivd.Name + "'"));
Type conditionalsType = typeof(List <>).MakeGenericType(marginalType);
IExpression conditionals_mpe = Builder.NewObject(conditionalsType);
VariableInformation conditionals_vi = VariableInformation.GetVariableInformation(context, vtc.conditionalsChannel.decl);
conditionals_vi.marginalPrototypeExpression = conditionals_mpe;
}
else
{
vtc.samplesChannel = vtc.marginalChannel;
vtc.conditionalsChannel = vtc.marginalChannel;
}
// Create uses channel
vtc.usageChannel = ChannelInfo.UseChannel(vi);
vtc.usageChannel.decl = vi.DeriveArrayVariable(stmts, context, vi.Name + "_uses", Builder.LiteralExpr(useCount), Builder.VarDecl("_ind", typeof(int)), useLiteralIndices: true);
context.InputAttributes.CopyObjectAttributesTo <InitialiseTo>(vi.declaration, context.OutputAttributes, vtc.usageChannel.decl);
context.OutputAttributes.Set(vtc.usageChannel.decl, vtc.usageChannel);
context.OutputAttributes.Set(vtc.usageChannel.decl, new DescriptionAttribute("uses of '" + ivd.Name + "'"));
SetMarginalPrototype(vtc.usageChannel.decl);
//setAllGroupRoots(context, ivd, false);
context.AddStatementsBeforeCurrent(stmts);
// Append usageDepth indices to def/marginal/use expressions
IExpression defExpr = Builder.VarRefExpr(ivd);
IExpression marginalExpr = Builder.VarRefExpr(vtc.marginalChannel.decl);
IExpression usageExpr = Builder.VarRefExpr(vtc.usageChannel.decl);
IExpression countExpr = Builder.LiteralExpr(useCount);
// Add clone factor tying together all of the channels
IMethodInvokeExpression usesEqualDefExpression;
Type[] genArgs = new Type[] { vi.varType };
if (algorithm is GibbsSampling && ((GibbsSampling)algorithm).UseSideChannels)
{
GibbsSampling gs = (GibbsSampling)algorithm;
IExpression burnInExpr = Builder.LiteralExpr(gs.BurnIn);
IExpression thinExpr = Builder.LiteralExpr(gs.Thin);
IExpression samplesExpr = Builder.VarRefExpr(vtc.samplesChannel.decl);
IExpression conditionalsExpr = Builder.VarRefExpr(vtc.conditionalsChannel.decl);
if (isDerived)
{
Delegate d = new FuncOut3 <PlaceHolder, int, int, int, PlaceHolder, PlaceHolder, PlaceHolder, PlaceHolder[]>(Factor.ReplicateWithMarginalGibbs);
usesEqualDefExpression = Builder.StaticGenericMethod(d, genArgs, defExpr, countExpr, burnInExpr, thinExpr, marginalExpr, samplesExpr, conditionalsExpr);
}
else
{
Delegate d = new FuncOut3 <PlaceHolder, int, int, int, PlaceHolder, PlaceHolder, PlaceHolder, PlaceHolder[]>(Factor.UsesEqualDefGibbs);
usesEqualDefExpression = Builder.StaticGenericMethod(d, genArgs, defExpr, countExpr, burnInExpr, thinExpr, marginalExpr, samplesExpr, conditionalsExpr);
}
}
else
{
Delegate d;
if (isDerived)
{
d = new FuncOut <PlaceHolder, int, PlaceHolder, PlaceHolder[]>(Factor.ReplicateWithMarginal <PlaceHolder>);
}
else
{
d = new FuncOut <PlaceHolder, int, PlaceHolder, PlaceHolder[]>(Factor.UsesEqualDef <PlaceHolder>);
}
usesEqualDefExpression = Builder.StaticGenericMethod(d, genArgs, defExpr, countExpr, marginalExpr);
}
if (isDerived)
{
context.OutputAttributes.Set(usesEqualDefExpression, new DerivedVariable()); // used by Gibbs
}
// Mark this as a pseudo-factor
context.OutputAttributes.Set(usesEqualDefExpression, new IsVariableFactor());
if (useCount == 1)
{
context.OutputAttributes.Set(usesEqualDefExpression, new DivideMessages(false));
}
else
{
context.InputAttributes.CopyObjectAttributesTo <DivideMessages>(ivd, context.OutputAttributes, usesEqualDefExpression);
}
context.InputAttributes.CopyObjectAttributesTo <GivePriorityTo>(ivd, context.OutputAttributes, usesEqualDefExpression);
IAssignExpression assignExpr = Builder.AssignExpr(usageExpr, usesEqualDefExpression);
// Copy attributes across from input to output
Context.InputAttributes.CopyObjectAttributesTo <Algorithm>(ivd, context.OutputAttributes, assignExpr);
context.OutputAttributes.Remove <InitialiseTo>(ivd);
if (vi.ArrayDepth == 0)
{
// Insert the UsesEqualDef statement after the declaration.
// Note the variable will not have been defined yet.
context.AddStatementAfterCurrent(Builder.ExprStatement(assignExpr));
}
else
{
// For an array, the UsesEqualDef statement should be inserted after the array is allocated.
// Store the statement for later use by ConvertArrayCreate.
context.InputAttributes.Remove <LoopContext>(ivd);
context.InputAttributes.Set(ivd, new LoopContext(context));
context.InputAttributes.Remove <Containers>(ivd);
context.InputAttributes.Set(ivd, new Containers(context));
vtc.usesEqualDefsStatements = Builder.StmtCollection();
vtc.usesEqualDefsStatements.Add(Builder.ExprStatement(assignExpr));
}
}
// These must be set after the above or they will be copied to the other channels
context.OutputAttributes.Set(ivd, vtc.defChannel);
context.OutputAttributes.Set(vtc.defChannel.decl, new DescriptionAttribute("definition of '" + ivd.Name + "'"));
return(ivde);
}
/// <summary>
/// Converts a variable declaration by creating definition, marginal and uses channel variables.
/// </summary>
protected override IExpression ConvertVariableDeclExpr(IVariableDeclarationExpression ivde)
{
IVariableDeclaration ivd = ivde.Variable;
VariableInformation vi = VariableInformation.GetVariableInformation(context, ivd);
// If the variable is deterministic, return
bool isInferred = context.InputAttributes.Has <IsInferred>(ivd);
if (!vi.IsStochastic)
{
return(ivde);
}
bool suppressVariableFactor = context.InputAttributes.Has <SuppressVariableFactor>(ivd);
bool isDerived = context.InputAttributes.Has <DerivedVariable>(ivd);
bool isStochastic = vi.IsStochastic;
int useCount;
ChannelAnalysisTransform.UsageInfo info;
if (!analysis.usageInfo.TryGetValue(ivd, out info))
{
useCount = 0;
}
else
{
useCount = info.NumberOfUsesOld;
}
if (!(algorithm is Algorithms.GibbsSampling) && isStochastic && !suppressVariableFactor && (useCount <= 1) && !isInferred && isDerived)
{
// this is optional
suppressVariableFactor = true;
context.InputAttributes.Set(ivd, new SuppressVariableFactor());
}
context.InputAttributes.Remove <LoopContext>(ivd);
context.InputAttributes.Set(ivd, new LoopContext(context));
// Create variable-to-channel information for the variable.
VariableToChannelInformation vtci = new VariableToChannelInformation();
vtci.shareAllUses = (useCount <= 1);
Context.InputAttributes.Set(ivd, vtci);
// Create the definition channel
vtci.defChannel = ChannelInfo.DefChannel(vi);
vtci.defChannel.decl = ivd;
// Always create a variable factor for a stochastic variable
if (isInferred || (isStochastic && !suppressVariableFactor))
{
vi.DefineAllIndexVars(context);
IList <IStatement> stmts = Builder.StmtCollection();
// Create marginal channel
CreateMarginalChannel(vi, vtci, stmts);
if (algorithm is GibbsSampling && ((GibbsSampling)algorithm).UseSideChannels)
{
CreateSamplesChannel(vi, vtci, stmts);
CreateConditionalsChannel(vi, vtci, stmts);
}
else
{
vtci.samplesChannel = vtci.marginalChannel;
vtci.conditionalsChannel = vtci.marginalChannel;
}
if (isStochastic)
{
// Create uses channel
CreateUsesChannel(vi, useCount, vtci, stmts);
}
//setAllGroupRoots(context, ivd, false);
context.AddStatementsBeforeCurrent(stmts);
// Append usageDepth indices to def/marginal/use expressions
IExpression defExpr = Builder.VarRefExpr(ivd);
IExpression marginalExpr = Builder.VarRefExpr(vtci.marginalChannel.decl);
IExpression countExpr = Builder.LiteralExpr(useCount);
// Add clone factor tying together all of the channels
IMethodInvokeExpression variableFactorExpr;
Type[] genArgs = new Type[] { vi.varType };
if (algorithm is GibbsSampling && ((GibbsSampling)algorithm).UseSideChannels)
{
GibbsSampling gs = (GibbsSampling)algorithm;
IExpression burnInExpr = Builder.LiteralExpr(gs.BurnIn);
IExpression thinExpr = Builder.LiteralExpr(gs.Thin);
IExpression samplesExpr = Builder.VarRefExpr(vtci.samplesChannel.decl);
IExpression conditionalsExpr = Builder.VarRefExpr(vtci.conditionalsChannel.decl);
if (isDerived)
{
Delegate d = new FuncOut3 <PlaceHolder, int, int, int, PlaceHolder, PlaceHolder, PlaceHolder, PlaceHolder[]>(Clone.ReplicateWithMarginalGibbs);
variableFactorExpr = Builder.StaticGenericMethod(d, genArgs, defExpr, countExpr, burnInExpr, thinExpr, marginalExpr, samplesExpr, conditionalsExpr);
}
else
{
Delegate d = new FuncOut3 <PlaceHolder, int, int, int, PlaceHolder, PlaceHolder, PlaceHolder, PlaceHolder[]>(Clone.UsesEqualDefGibbs);
variableFactorExpr = Builder.StaticGenericMethod(d, genArgs, defExpr, countExpr, burnInExpr, thinExpr, marginalExpr, samplesExpr, conditionalsExpr);
}
}
else
{
Delegate d;
if (isDerived)
{
d = new FuncOut <PlaceHolder, int, PlaceHolder, PlaceHolder[]>(Clone.ReplicateWithMarginal <PlaceHolder>);
}
else
{
d = new FuncOut <PlaceHolder, int, PlaceHolder, PlaceHolder[]>(Clone.UsesEqualDef <PlaceHolder>);
}
variableFactorExpr = Builder.StaticGenericMethod(d, genArgs, defExpr, countExpr, marginalExpr);
}
if (isDerived)
{
context.OutputAttributes.Set(variableFactorExpr, new DerivedVariable()); // used by Gibbs
}
// Mark this as a pseudo-factor
context.OutputAttributes.Set(variableFactorExpr, new IsVariableFactor());
if (useCount <= 1)
{
context.OutputAttributes.Set(variableFactorExpr, new DivideMessages(false));
}
else
{
context.InputAttributes.CopyObjectAttributesTo <DivideMessages>(ivd, context.OutputAttributes, variableFactorExpr);
}
context.InputAttributes.CopyObjectAttributesTo <GivePriorityTo>(ivd, context.OutputAttributes, variableFactorExpr);
if (vtci.usageChannel != null)
{
IExpression usageExpr = Builder.VarRefExpr(vtci.usageChannel.decl);
IAssignExpression assignExpr = Builder.AssignExpr(usageExpr, variableFactorExpr);
// Copy attributes across from input to output
Context.InputAttributes.CopyObjectAttributesTo <Algorithm>(ivd, context.OutputAttributes, assignExpr);
context.OutputAttributes.Remove <InitialiseTo>(ivd);
if (vi.ArrayDepth == 0)
{
// Insert the UsesEqualDef statement after the declaration.
// Note the variable will not have been defined yet.
context.AddStatementAfterCurrent(Builder.ExprStatement(assignExpr));
}
else
{
// For an array, the UsesEqualDef statement should be inserted after the array is allocated.
// Store the statement for later use by ConvertArrayCreate.
context.InputAttributes.Remove <LoopContext>(ivd);
context.InputAttributes.Set(ivd, new LoopContext(context));
context.InputAttributes.Remove <Containers>(ivd);
context.InputAttributes.Set(ivd, new Containers(context));
vtci.usesEqualDefsStatements = Builder.StmtCollection();
vtci.usesEqualDefsStatements.Add(Builder.ExprStatement(assignExpr));
}
}
}
// These must be set after the above or they will be copied to the other channels
if (isStochastic)
{
context.OutputAttributes.Set(ivd, vtci.defChannel);
}
if (!context.InputAttributes.Has <DescriptionAttribute>(vtci.defChannel.decl))
{
context.OutputAttributes.Set(vtci.defChannel.decl, new DescriptionAttribute("definition of '" + ivd.Name + "'"));
}
return(ivde);
}
private void BugsRats(bool initialiseAlpha, bool initialiseAlphaC)
{
Rand.Restart(0);
double precOfGaussianPrior = 1.0E-6;
double shapeRateOfGammaPrior = 0.02; // smallest choice that will avoid zeros
double meanOfBetaPrior = 0.0;
double meanOfAlphaPrior = 0.0;
// The model
int N = RatsHeightData.GetLength(0);
int T = RatsHeightData.GetLength(1);
double xbar = 22.0;
double[] xDataZeroMean = new double[RatsXData.Length];
for (int i = 0; i < RatsXData.Length; i++)
{
xDataZeroMean[i] = RatsXData[i] - xbar;
}
Range r = new Range(N).Named("N");
Range w = new Range(T).Named("T");
VariableArray2D <double> y = Variable.Observed <double>(RatsHeightData, r, w).Named("y");
VariableArray <double> x = Variable.Observed <double>(xDataZeroMean, w).Named("x");
Variable <double> tauC = Variable.GammaFromShapeAndRate(shapeRateOfGammaPrior, shapeRateOfGammaPrior).Named("tauC");
Variable <double> alphaC = Variable.GaussianFromMeanAndPrecision(meanOfAlphaPrior, precOfGaussianPrior).Named("alphaC");
Variable <double> alphaTau = Variable.GammaFromShapeAndRate(shapeRateOfGammaPrior, shapeRateOfGammaPrior).Named("alphaTau");
Variable <double> betaC = Variable.GaussianFromMeanAndPrecision(meanOfBetaPrior, precOfGaussianPrior).Named("betaC");
Variable <double> betaTau = Variable.GammaFromShapeAndRate(shapeRateOfGammaPrior, shapeRateOfGammaPrior).Named("betaTau");
VariableArray <double> alpha = Variable.Array <double>(r).Named("alpha");
alpha[r] = Variable.GaussianFromMeanAndPrecision(alphaC, alphaTau).ForEach(r);
VariableArray <double> beta = Variable.Array <double>(r).Named("beta");
beta[r] = Variable.GaussianFromMeanAndPrecision(betaC, betaTau).ForEach(r);
VariableArray2D <double> mu = Variable.Array <double>(r, w).Named("mu");
VariableArray2D <double> betaX = Variable.Array <double>(r, w).Named("betax");
betaX[r, w] = beta[r] * x[w];
mu[r, w] = alpha[r] + betaX[r, w];
y[r, w] = Variable.GaussianFromMeanAndPrecision(mu[r, w], tauC);
Variable <double> alpha0 = (alphaC - xbar * betaC).Named("alpha0");
InferenceEngine ie;
GibbsSampling gs = new GibbsSampling();
// Initialise both alpha and beta together.
// Initialising only alpha (or only beta) is not reliable because you could by chance get a large betaTau and small tauC to start,
// at which point beta and alphaC become garbage, leading to alpha becoming garbage on the next iteration.
bool initialiseBeta = initialiseAlpha;
bool initialiseBetaC = initialiseAlphaC;
if (initialiseAlpha)
{
Gaussian[] alphaInit = new Gaussian[N];
for (int i = 0; i < N; i++)
{
alphaInit[i] = Gaussian.FromMeanAndPrecision(250.0, 1.0);
}
alpha.InitialiseTo(Distribution <double> .Array(alphaInit));
}
if (initialiseBeta)
{
Gaussian[] betaInit = new Gaussian[N];
for (int i = 0; i < N; i++)
{
betaInit[i] = Gaussian.FromMeanAndPrecision(6.0, 1.0);
}
beta.InitialiseTo(Distribution <double> .Array(betaInit));
}
if (initialiseAlphaC)
{
alphaC.InitialiseTo(Gaussian.FromMeanAndVariance(250.0, 1.0));
}
if (initialiseBetaC)
{
betaC.InitialiseTo(Gaussian.FromMeanAndVariance(6.0, 1.0));
}
if (false)
{
//tauC.InitialiseTo(Gamma.FromMeanAndVariance(1.0, 0.1));
//alphaTau.InitialiseTo(Gamma.FromMeanAndVariance(1.0, 0.1));
//betaTau.InitialiseTo(Gamma.FromMeanAndVariance(1.0, 0.1));
}
if (!initialiseAlpha && !initialiseBeta && !initialiseAlphaC && !initialiseBetaC)
{
gs.BurnIn = 1000;
}
ie = new InferenceEngine(gs);
ie.ShowProgress = false;
ie.ModelName = "BugsRats";
ie.NumberOfIterations = 4000;
ie.OptimiseForVariables = new List <IVariable>()
{
alphaC, betaC, alpha0, tauC
};
betaC.AddAttribute(QueryTypes.Marginal);
betaC.AddAttribute(QueryTypes.Samples);
alpha0.AddAttribute(QueryTypes.Marginal);
alpha0.AddAttribute(QueryTypes.Samples);
tauC.AddAttribute(QueryTypes.Marginal);
tauC.AddAttribute(QueryTypes.Samples);
// Inference
object alphaCActual = ie.Infer(alphaC);
Gaussian betaCMarg = ie.Infer <Gaussian>(betaC);
Gaussian alpha0Marg = ie.Infer <Gaussian>(alpha0);
Gamma tauCMarg = ie.Infer <Gamma>(tauC);
// Check results against BUGS
Gaussian betaCExpected = new Gaussian(6.185, System.Math.Pow(0.1068, 2));
Gaussian alpha0Expected = new Gaussian(106.6, System.Math.Pow(3.625, 2));
double sigmaMeanExpected = 6.082;
double sigmaMean = System.Math.Sqrt(1.0 / tauCMarg.GetMean());
if (!initialiseAlpha && !initialiseAlphaC)
{
Debug.WriteLine("betaC = {0} should be {1}", betaCMarg, betaCExpected);
Debug.WriteLine("alpha0 = {0} should be {1}", alpha0Marg, alpha0Expected);
}
Assert.True(GaussianDiff(betaCExpected, betaCMarg) < 0.1);
Assert.True(GaussianDiff(alpha0Expected, alpha0Marg) < 0.1);
Assert.True(MMath.AbsDiff(sigmaMeanExpected, sigmaMean, 0.1) < 0.1);
IList <double> betaCSamples = ie.Infer <IList <double> >(betaC, QueryTypes.Samples);
IList <double> alpha0Samples = ie.Infer <IList <double> >(alpha0, QueryTypes.Samples);
IList <double> tauCSamples = ie.Infer <IList <double> >(tauC, QueryTypes.Samples);
GaussianEstimator est = new GaussianEstimator();
foreach (double sample in betaCSamples)
{
est.Add(sample);
}
Gaussian betaCMarg2 = est.GetDistribution(new Gaussian());
Assert.True(GaussianDiff(betaCMarg, betaCMarg2) < 0.1);
}
