[Insight-users] Fwd: KMeans type mismatch

[Luis Ibanez]

Hi David,

The Parameters Array expected by the Estimator
is a linear array of doubles.


Therefore, instead of the code:

       unsigned int NumClasses = 3;
       EstimatorType::ParametersType initialMeans(NumClasses);
       initialMeans[0] = InitialMean0;
       initialMeans[1] = InitialMean1;
       initialMeans[2] = InitialMean2;
       estimator->SetParameters( initialMeans );


You should do:

       const unsigned int NumClasses = 3;
       const unsigned int VectorLength = 3;
       EstimatorType::ParametersType initialMeans(NumClasses *
VectorLength);
       initialMeans[0] = InitialMean0[0];
       initialMeans[1] = InitialMean0[1];
       initialMeans[2] = InitialMean0[2];
       initialMeans[3] = InitialMean1[0];
       initialMeans[4] = InitialMean1[1];
       initialMeans[5] = InitialMean1[2];
       initialMeans[6] = InitialMean2[0];
       initialMeans[7] = InitialMean2[1];
       initialMeans[8] = InitialMean2[2];
       estimator->SetParameters( initialMeans );


Regards,


       Luis



---------------------------------------------
On Tue, Jul 28, 2009 at 7:52 AM, David Doria
<daviddoria+itk at gmail.com<daviddoria%2Bitk at gmail.com>
> wrote:

> ---------- Forwarded message ----------
> From: David Doria <daviddoria+itk at gmail.com <daviddoria%2Bitk at gmail.com>>
> Date: Thu, Jul 23, 2009 at 9:18 AM
> Subject: KMeans type mismatch
> To: ITK <insight-users at itk.org>
>
>
> I want to do a KMeans clustering of 3d points. So I setup a 3-vector like
> this:
>
>        typedef itk::Vector< double, 3 > MeasurementVectorType;
>
> and fill it with some points that are clearly in 3 clusters:
>
>        MeasurementVectorType p0, p1, p2, p3, p4, p5, p6, p7, p8;
>
>        //cluster 1
>        p0[0]=  0.0; p0[1]= 0.0; p0[2]= 0.0;
>        p1[0]=  0.1; p1[1]= 0.0; p1[2]= 0.0;
>        p2[0]=  0.0; p2[1]= 0.1; p2[2]= 0.0;
>
>        //cluster 2
>        p3[0]=  5.0; p3[1]=  5.0; p3[2]= 5.0;
>        p4[0]=  5.1; p4[1]=  5.0; p4[2]= 5.0;
>        p5[0]=  5.0; p5[1]=  5.1; p5[2]= 5.0;
>
>        //cluster 3
>        p6[0]=  -5.0; p6[1]=  -5.0; p6[2]= -5.0;
>        p7[0]=  -5.1; p7[1]=  -5.0; p7[2]= -5.0;
>        p8[0]=  -5.0; p8[1]=  -5.1; p8[2]= -5.0;
>
> Then I setup the ListSample for the KdTree:
>
>        typedef itk::Statistics::ListSample< MeasurementVectorType >
> SampleType;
>        SampleType::Pointer sample = SampleType::New();
>        sample->PushBack(p0);
>        sample->PushBack(p1);
>        sample->PushBack(p2);
>        sample->PushBack(p3);
>        sample->PushBack(p4);
>        sample->PushBack(p5);
>        sample->PushBack(p6);
>        sample->PushBack(p7);
>        sample->PushBack(p8);
>
>        typedef itk::Statistics::WeightedCentroidKdTreeGenerator< SampleType
> > TreeGeneratorType;
>        //typedef itk::Statistics::KdTreeGenerator< SampleType >
> TreeGeneratorType;
>        TreeGeneratorType::Pointer treeGenerator = TreeGeneratorType::New();
>        treeGenerator->SetSample( sample );
>        treeGenerator->SetBucketSize( 3 ); //number of measurement vectors
> in
> a terminal node
>        treeGenerator->Update();
>
> Now I want to seed the KMeans algorithm with 3 points in the 3d space:
>        // Once we have the k-d tree, it is a simple procedure to produce k
> mean estimates
>
>        typedef TreeGeneratorType::KdTreeType TreeType;
>        typedef itk::Statistics::KdTreeBasedKmeansEstimator<TreeType>
> EstimatorType;
>        EstimatorType::Pointer estimator = EstimatorType::New();
>
>        MeasurementVectorType InitialMean0, InitialMean1, InitialMean2;
>        InitialMean0[0] = 1.0;
>        InitialMean0[1] = 1.0;
>        InitialMean0[2] = 1.0;
>
>        InitialMean1[0] = -1.0;
>        InitialMean1[1] = -1.0;
>        InitialMean1[2] = -1.0;
>
>        InitialMean2[0] = 0.0;
>        InitialMean2[1] = 0.0;
>        InitialMean2[2] = 0.0;
>
> I create a vector of 3d vectors:
>        unsigned int NumClasses = 3;
>        EstimatorType::ParametersType initialMeans(NumClasses);
>        initialMeans[0] = InitialMean0;
>        initialMeans[1] = InitialMean1;
>        initialMeans[2] = InitialMean2;
>
> In hopes to do something like this:
>
>        estimator->SetParameters( initialMeans );
>        estimator->SetKdTree( treeGenerator->GetOutput() );
>        estimator->SetMaximumIteration( 200 );
>        estimator->SetCentroidPositionChangesThreshold(0.0);
>        estimator->StartOptimization();
>
> But it complains on these lines:
>        initialMeans[0] = InitialMean0;
>
> error: cannot convert 'main()::MeasurementVectorType' to 'double' in
> assignment
>
> How would I setup these initial means?
>
> On a separate note: is there an agglomerative clustering algorithm in
> ITK? That is, have it start with the 9 points in a single cluster and
> then split the clusters until some "continue splitting" tolerance is
> met?
>
> Thanks,
>
> David
>
> ----------------
> Anyone know how to fix this type issue?
>
> Thanks,
>
> David
>
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20090728/b3a3c1c9/attachment.htm>