#include "daal.h"
#include "service.h"
using namespace std;
using namespace daal;
using namespace daal::algorithms;
string trainedModelsFileNames[] = { "../data/batch/svm_multi_class_trained_model_01.csv",
"../data/batch/svm_multi_class_trained_model_02.csv",
"../data/batch/svm_multi_class_trained_model_12.csv" };
float biases[] = {-0.774F, -1.507F, -7.559F};
string testDatasetFileName = "../data/batch/multiclass_iris_train.csv";
const size_t nFeatures = 4;
const size_t nClasses = 3;
services::SharedPtr<svm::prediction::Batch<> > prediction(new svm::prediction::Batch<>());
classifier::prediction::ResultPtr predictionResult;
kernel_function::KernelIfacePtr kernel(new kernel_function::linear::Batch<>());
NumericTablePtr testGroundTruth;
multi_class_classifier::ModelPtr buildModelFromTraining();
void testModel(multi_class_classifier::ModelPtr& inputModel);
int main(int argc, char *argv[])
{
checkArguments(argc, argv, 1, &testDatasetFileName);
multi_class_classifier::ModelPtr builtModel = buildModelFromTraining();
prediction->parameter.kernel = kernel;
testModel(builtModel);
return 0;
}
multi_class_classifier::ModelPtr buildModelFromTraining()
{
multi_class_classifier::ModelBuilder<> multiBuilder(nFeatures, nClasses);
size_t imodel = 0;
for (size_t iClass = 1; iClass < nClasses; iClass++)
{
for (size_t jClass = 0; jClass < iClass; jClass++, imodel++)
{
FileDataSource<CSVFeatureManager> modelSource(trainedModelsFileNames[imodel],
DataSource::doAllocateNumericTable,
DataSource::doDictionaryFromContext);
NumericTablePtr supportVectors(new HomogenNumericTable<>(nFeatures, 0, NumericTable::doNotAllocate));
NumericTablePtr classificationCoefficients(new HomogenNumericTable<>(1, 0, NumericTable::doNotAllocate));
NumericTablePtr mergedModel(new MergedNumericTable(supportVectors, classificationCoefficients));
modelSource.loadDataBlock(mergedModel.get());
float bias = biases[imodel];
size_t nSV = supportVectors->getNumberOfRows();
BlockDescriptor<> blockResult;
supportVectors->getBlockOfRows(0, nSV, readOnly, blockResult);
float* first = blockResult.getBlockPtr();
float* last = first + nSV*nFeatures;
svm::ModelBuilder<> modelBuilder(nFeatures, nSV);
modelBuilder.setSupportVectors(first, last);
supportVectors->releaseBlockOfRows(blockResult);
classificationCoefficients->getBlockOfRows(0, nSV, readOnly, blockResult);
first = blockResult.getBlockPtr();
last = first + nSV;
modelBuilder.setClassificationCoefficients(first, last);
classificationCoefficients->releaseBlockOfRows(blockResult);
modelBuilder.setBias(bias);
multiBuilder.setTwoClassClassifierModel(jClass,iClass,modelBuilder.getModel());
}
}
return multiBuilder.getModel();
}
void testModel(multi_class_classifier::ModelPtr& inputModel)
{
FileDataSource<CSVFeatureManager> testDataSource(testDatasetFileName,
DataSource::doAllocateNumericTable,
DataSource::doDictionaryFromContext);
NumericTablePtr testData(new HomogenNumericTable<>(nFeatures, 0, NumericTable::doNotAllocate));
testGroundTruth = NumericTablePtr(new HomogenNumericTable<>(1, 0, NumericTable::doNotAllocate));
NumericTablePtr mergedData(new MergedNumericTable(testData, testGroundTruth));
testDataSource.loadDataBlock(mergedData.get());
multi_class_classifier::prediction::Batch<> algorithm(nClasses);
algorithm.parameter.prediction = prediction;
algorithm.input.set(classifier::prediction::data, testData);
algorithm.input.set(classifier::prediction::model, inputModel);
algorithm.compute();
predictionResult = algorithm.getResult();
printNumericTables<int, int>(testGroundTruth,
predictionResult->get(classifier::prediction::prediction),
"Ground truth", "Classification results",
"Multi-class SVM classification sample program results (first 20 observations):", 20);
}