Remove From My Forums

switch should be placed outside 'for' loop

Question

I am continuing building the Markov Mixture model. I realized that for the start state, I want to keep a single Variable.Discrete variable that will be shared by all sequences. When I make this change, I get an error message:

Compiling model...compilation failed.

Unhandled Exception:
MicrosoftResearch.Infer.CompilationFailedException: GateTransform failed with 1 error(s) and 0 warning(s):
Error 0: 'switch(ZeroState)' should be placed outside 'for(N)' in
if (ZeroState==K) {
	States[N][0] = Factor.Discrete(CPTTrans[K]);
}


  at MicrosoftResearch.Infer.ModelCompiler.GetTransformedDeclaration (MicrosoftResearch.Transforms.CodeModel.Interfaces.ITypeDeclaration itd, System.Reflection.MethodBase method, MicrosoftResearch.Transforms.AttributeRegistry`2[TObject,TAttribute] inputAttributes) [0x00155] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.ModelCompiler.CompileWithoutParams (MicrosoftResearch.Transforms.CodeModel.Interfaces.ITypeDeclaration itd, System.Reflection.MethodBase method, MicrosoftResearch.Transforms.AttributeRegistry`2[TObject,TAttribute] inputAttributes) [0x0000c] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.Compile () [0x000b5] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.BuildAndCompile (System.Boolean inferOnlySpecifiedVars, System.Collections.Generic.IEnumerable`1[T] vars) [0x000e1] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.GetCompiledInferenceAlgorithm (System.Boolean inferOnlySpecifiedVars, MicrosoftResearch.Infer.Models.IVariable var) [0x00017] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.InferAll (System.Boolean inferOnlySpecifiedVars, MicrosoftResearch.Infer.Models.IVariable var) [0x00000] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.Infer[TReturn] (MicrosoftResearch.Infer.Models.IVariable var) [0x00000] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MarkovMixtureModel.MarkovMixtureModel.InferPosteriors () [0x00001] in /Users/vlad/Projects/MarkovMixtureModel/MarkovMixtureModel/MarkovMixtureModel.cs:110 
  at MarkovMixtureModel.Program.TestMarkovMixtureModel () [0x00104] in /Users/vlad/Projects/MarkovMixtureModel/MarkovMixtureModel/Program.cs:51 
  at MarkovMixtureModel.Program.Main (System.String[] args) [0x00001] in /Users/vlad/Projects/MarkovMixtureModel/MarkovMixtureModel/Program.cs:15 
[ERROR] FATAL UNHANDLED EXCEPTION: MicrosoftResearch.Infer.CompilationFailedException: GateTransform failed with 1 error(s) and 0 warning(s):
Error 0: 'switch(ZeroState)' should be placed outside 'for(N)' in
if (ZeroState==K) {
	States[N][0] = Factor.Discrete(CPTTrans[K]);
}


  at MicrosoftResearch.Infer.ModelCompiler.GetTransformedDeclaration (MicrosoftResearch.Transforms.CodeModel.Interfaces.ITypeDeclaration itd, System.Reflection.MethodBase method, MicrosoftResearch.Transforms.AttributeRegistry`2[TObject,TAttribute] inputAttributes) [0x00155] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.ModelCompiler.CompileWithoutParams (MicrosoftResearch.Transforms.CodeModel.Interfaces.ITypeDeclaration itd, System.Reflection.MethodBase method, MicrosoftResearch.Transforms.AttributeRegistry`2[TObject,TAttribute] inputAttributes) [0x0000c] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.Compile () [0x000b5] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.BuildAndCompile (System.Boolean inferOnlySpecifiedVars, System.Collections.Generic.IEnumerable`1[T] vars) [0x000e1] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.GetCompiledInferenceAlgorithm (System.Boolean inferOnlySpecifiedVars, MicrosoftResearch.Infer.Models.IVariable var) [0x00017] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.InferAll (System.Boolean inferOnlySpecifiedVars, MicrosoftResearch.Infer.Models.IVariable var) [0x00000] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MicrosoftResearch.Infer.InferenceEngine.Infer[TReturn] (MicrosoftResearch.Infer.Models.IVariable var) [0x00000] in <1c09b726d63a41118a8cecdae81e3110>:0 
  at MarkovMixtureModel.MarkovMixtureModel.InferPosteriors () [0x00001] in /Users/vlad/Projects/MarkovMixtureModel/MarkovMixtureModel/MarkovMixtureModel.cs:110 
  at MarkovMixtureModel.Program.TestMarkovMixtureModel () [0x00104] in /Users/vlad/Projects/MarkovMixtureModel/MarkovMixtureModel/Program.cs:51 
  at MarkovMixtureModel.Program.Main (System.String[] args) [0x00001] in /Users/vlad/Projects/MarkovMixtureModel/MarkovMixtureModel/Program.cs:15

My model currently is:

// define primary model variables -- zero state
ZeroState = Variable.Discrete(ProbInit).Named("ZeroState");

// define primary model variables -- actual states
States = Variable.Array(Variable.Array<int>(T), N).Named("States");

using (Variable.ForEach(N))
{
  using (var block = Variable.ForEach(T))
  {
    var t = block.Index;
    var previousState = States[N][t - 1];

    using (Variable.If(t == 0))
    {
       using (Variable.Switch(ZeroState))
         States[N][T].SetTo(Variable.Discrete(CPTTrans[ZeroState]));
    }

    using (Variable.If(t > 0))
    {
      using (Variable.Switch(previousState))
        States[N][T].SetTo(Variable.Discrete(CPTTrans[previousState]));
    }
  }
}

I have tried to put .ForEach(N) at relevant places, also the Variable.ForEach(N) blocks. To no avail. It looks like I am stuck.

Tuesday, March 6, 2018 8:01 AM

Answers

            using (Variable.Switch(ZeroState))
            using (Variable.ForEach(N))
            {
                using (var block = Variable.ForEach(T))
                {
                    var t = block.Index;

                    using (Variable.If(t == 0))
                    {
                        States[N][T].SetTo(Variable.Discrete(CPTTrans[ZeroState]));
                    }
                }
            }

            using (Variable.ForEach(N))
            {
                using (var block = Variable.ForEach(T))
                {
                    var t = block.Index;
                    var previousState = States[N][t - 1];

                    using (Variable.If(t > 0))
                    {
                        using (Variable.Switch(previousState))
                            States[N][T].SetTo(Variable.Discrete(CPTTrans[previousState]));
                    }
                }
            }

Marked as answer by usptact Tuesday, March 6, 2018 5:45 PM

Tuesday, March 6, 2018 2:47 PM

Owner

All replies

            using (Variable.Switch(ZeroState))
            using (Variable.ForEach(N))
            {
                using (var block = Variable.ForEach(T))
                {
                    var t = block.Index;

                    using (Variable.If(t == 0))
                    {
                        States[N][T].SetTo(Variable.Discrete(CPTTrans[ZeroState]));
                    }
                }
            }

            using (Variable.ForEach(N))
            {
                using (var block = Variable.ForEach(T))
                {
                    var t = block.Index;
                    var previousState = States[N][t - 1];

                    using (Variable.If(t > 0))
                    {
                        using (Variable.Switch(previousState))
                            States[N][T].SetTo(Variable.Discrete(CPTTrans[previousState]));
                    }
                }
            }

Marked as answer by usptact Tuesday, March 6, 2018 5:45 PM

Tuesday, March 6, 2018 2:47 PM

Owner

Thank you, Tom! I found the issue with my model. The initial states are modeled using "ProbInit", the transition matrix is not used. I was not careful enough.

The model now looks like this:

using (Variable.ForEach(N))
{
  using (var block = Variable.ForEach(T))
  {
    var t = block.Index;

    using (Variable.If(t == 0))
      States[N][T].SetTo(Variable.Discrete(ProbInit));

    var previousState = States[N][t - 1];

    using (Variable.If(t > 0))
      using (Variable.Switch(previousState))
        States[N][T].SetTo(Variable.Discrete(CPTTrans[previousState]));
  }
}

All model parameters are getting recovered successfully. Now to the next step - add clusters.

Tuesday, March 6, 2018 5:12 PM

Answered by:

switch should be placed outside 'for' loop

Question

Answers

All replies