Java Code Examples for org.neuroph.core.data.DataSet#getOutputSize()

/**
        * Creates training set for the specified image labels and rgb data
        * @param imageLabels image labels
        * @param rgbDataMap map collection of rgb data
        * @return training set for the specified image data
        */
public static DataSet createRGBTrainingSet(List<String> imageLabels, Map<String, FractionRgbData> rgbDataMap) 	{
               int inputCount = rgbDataMap.values().iterator().next().getFlattenedRgbValues().length;
               int outputCount = imageLabels.size();
	DataSet trainingSet = new DataSet(inputCount, outputCount);

	for (Entry<String, FractionRgbData> entry : rgbDataMap.entrySet()) {
		double[] input = entry.getValue().getFlattenedRgbValues();
		double[] response = createResponse(entry.getKey(), imageLabels);
		trainingSet.add(new DataSetRow(input, response));
	}

               // set labels for output columns
               int inputSize = trainingSet.getInputSize();
               for (int c= 0; c<trainingSet.getOutputSize() ; c++) {
                   trainingSet.setColumnName(inputSize+c, imageLabels.get(c));
               }

               return trainingSet;
}
 

/**
        * Creates training set for the specified image labels and hsl data
        * @param imageLabels image labels
        * @param hslDataMap map colletction of hsl data
        * @return training set for the specified image data
        */
       public static DataSet createHSLTrainingSet(List<String> imageLabels, Map<String, FractionHSLData> hslDataMap) 	{
               int inputCount = hslDataMap.values().iterator().next().getFlattenedHSLValues().length;
               int outputCount = imageLabels.size();
	DataSet trainingSet = new DataSet(inputCount, outputCount);

	for (Entry<String, FractionHSLData> entry : hslDataMap.entrySet()) {
		double[] input = entry.getValue().getFlattenedHSLValues();
		double[] response = createResponse(entry.getKey(), imageLabels);
		trainingSet.add(new DataSetRow(input, response));
	}

               // set labels for output columns
               int inputSize = trainingSet.getInputSize();
               for (int c= 0; c<trainingSet.getOutputSize() ; c++) {
                   trainingSet.setColumnName(inputSize+c, imageLabels.get(c));
               }


               return trainingSet;
}
 

/**
        * Creates binary black and white training set for the specified image labels and rgb data
        * white = 0 black = 1
        * @param imageLabels image labels
        * @param rgbDataMap map collection of rgb data
        * @return binary black and white training set for the specified image data
        */
       public static DataSet createBlackAndWhiteTrainingSet(List<String> imageLabels, Map<String, FractionRgbData> rgbDataMap) throws VectorSizeMismatchException
{
           // TODO: Use some binarization image filter to do this; currently it works  with averaging RGB values
               int inputCount = rgbDataMap.values().iterator().next().getFlattenedRgbValues().length / 3;
               int outputCount = imageLabels.size();
	DataSet trainingSet = new DataSet(inputCount, outputCount);

	for (Entry<String, FractionRgbData> entry : rgbDataMap.entrySet()) {
		double[] inputRGB = entry.getValue().getFlattenedRgbValues();
                       double[] inputBW = FractionRgbData.convertRgbInputToBinaryBlackAndWhite(inputRGB);
                       double[] response = createResponse(entry.getKey(), imageLabels);
		trainingSet.add(new DataSetRow(inputBW, response));
	}

               // set labels for output columns
               int inputSize = trainingSet.getInputSize();
               for (int c= 0; c<trainingSet.getOutputSize() ; c++) {
                   trainingSet.setColumnName(inputSize+c, imageLabels.get(c));
               }

           return trainingSet;
}
 

public ConfusionMatrix sumConfusionMatrix(List<ConfusionMatrix> cmList, DataSet dataSet) {
    ConfusionMatrix cm = new ConfusionMatrix(cmList.get(0).getClassLabels());
    int[][] ar = new int[dataSet.getOutputSize()][dataSet.getOutputSize()];
    for (ConfusionMatrix c : cmList) {
        for (int i = 0; i < dataSet.getOutputSize(); ++i) {
            for (int j = 0; j < dataSet.getOutputSize(); ++j) {
                ar[i][j] += c.get(i, j);
            }
        }
    }

    cm.setValues(ar);
    return cm;
}
 

public DataSetStatistics(DataSet dataSet) {
    this.dataSet = dataSet;
    this.rowLength = dataSet.getInputSize() + dataSet.getOutputSize();
    this.rowsCount = dataSet.getRows().size();
    this.mean = new double[this.rowLength];
    this.max = new double[this.rowLength];
    this.min = new double[this.rowLength];
    this.sum = new double[this.rowLength];
    this.variance = new double[this.rowLength];
    this.stdDeviation = new double[this.rowLength];
    this.frequency = new double[this.rowLength];
    this.setDefaultValues();
}
 

/**
 * Finds max values for all columns in dataset (inputs and outputs)
 * Sets max column values to maxIn and maxOut fields
 * @param dataSet
 */
private void findMaxVectors(DataSet dataSet) {
    int inputSize = dataSet.getInputSize();
    int outputSize = dataSet.getOutputSize();

    // init with minimum values
    maxIn = new double[inputSize];
    for (int i = 0; i < inputSize; i++) {
        maxIn[i] = Double.MIN_VALUE;
    }

    maxOut = new double[outputSize];
    for (int i = 0; i < outputSize; i++) {
        maxOut[i] = Double.MIN_VALUE;
    }


    for (DataSetRow dataSetRow : dataSet.getRows()) {
        double[] input = dataSetRow.getInput();
        for (int i = 0; i < inputSize; i++) {
            if (input[i] > maxIn[i]) {
                maxIn[i] = input[i];
            }
        }

        double[] output = dataSetRow.getDesiredOutput();
        for (int i = 0; i < outputSize; i++) {
            if (output[i] > maxOut[i]) {
                maxOut[i] = output[i];
            }
        }

    }
}
 

/**
* Finds max values for columns in input and output vector for given data set
* @param dataSet
*/
private void init(DataSet dataSet) {
    int inputSize = dataSet.getInputSize();
    int outputSize = dataSet.getOutputSize();

    maxIn = new double[inputSize];
    for(int i=0; i<inputSize; i++) {
        maxIn[i] = Double.MIN_VALUE;
    }

    maxOut = new double[outputSize];
    for(int i=0; i<outputSize; i++)
        maxOut[i] = Double.MIN_VALUE;

    for (DataSetRow dataSetRow : dataSet.getRows()) {
        double[] input = dataSetRow.getInput();
        for (int i = 0; i < inputSize; i++) {
            if (Math.abs(input[i]) > maxIn[i]) {
                maxIn[i] = Math.abs(input[i]);
            }
         }

        double[] output = dataSetRow.getDesiredOutput();
        for (int i = 0; i < outputSize; i++) {
            if (Math.abs(output[i]) > maxOut[i]) {
                maxOut[i] = Math.abs(output[i]);
            }
        }
    }
}
 

@Override
public DataSet[] sample(DataSet dataSet) {
    // if object was initializes by specifying numParts calculate subSetSizes so all subsets are equally sized
    if (subSetSizes == null) {
        final double singleSubSetSize = 1.0d / numSubSets;
        subSetSizes = new double[numSubSets];
        for (int i = 0; i < numSubSets; i++) {
            subSetSizes[i] = singleSubSetSize;
        }
    }

    // create list of data sets to return
    List<DataSet> subSets = new ArrayList<>();

    // shuffle dataset in order to randomize rows that will be used to fill subsets
    dataSet.shuffle();

    int idxCounter = 0; // index of main data set
    for (int subSetIdx = 0; subSetIdx < numSubSets; subSetIdx++) {
        // create new subset
        DataSet newSubSet = new DataSet(dataSet.getInputSize(), dataSet.getOutputSize());
        // cop column names if there are any
        newSubSet.setColumnNames(dataSet.getColumnNames());

        // fill subset with rows
        long subSetSize = Math.round(subSetSizes[subSetIdx] * dataSet.size()); // calculate size of the current subset
        for (int i = 0; i < subSetSize; i++) {
            if (idxCounter >= dataSet.size()) {
                break;
            }
            newSubSet.add(dataSet.getRowAt(idxCounter));
            idxCounter++;
        }

        // add current subset to list that will be returned
        subSets.add(newSubSet);
    }

    return subSets.toArray(new DataSet[numSubSets]);
}
 

private DataSet createDataSetFold(DataSet dataSet) {

        DataSet foldSet = new DataSet(dataSet.getInputSize(), dataSet.getOutputSize());
        for (int j = 0; j < getSampleSize(); j++) {
            foldSet.add(getNextDataSetRow());
        }

        return foldSet;
    }
 

/**
 * Create instance of training with new neural network
 *
 * @param dataset
 * @param settings
 */
public AbstractTraining(DataSet dataset, TrainingSettings settings) {
    this.dataset = dataset;
    this.settings = settings;
    this.stats = new TrainingStatistics();
    this.neuralNet = new MultiLayerPerceptron(TransferFunctionType.SIGMOID, dataset.getInputSize(), settings.getHiddenNeurons(), dataset.getOutputSize());
}
 

/**
 * Find min and max values for each position in vectors.
 *
 * @param dataSet
 */
private void findMaxAndMinVectors(DataSet dataSet) {
    int inputSize = dataSet.getInputSize();
    int outputSize = dataSet.getOutputSize();

    maxIn = new double[inputSize];
    minIn = new double[inputSize];

    for(int i=0; i<inputSize; i++) {
        maxIn[i] = Double.MIN_VALUE;
        minIn[i] = Double.MAX_VALUE;
    }

    maxOut = new double[outputSize];
    minOut = new double[outputSize];

    for(int i=0; i<outputSize; i++) {
        maxOut[i] = Double.MIN_VALUE;
        minOut[i] = Double.MAX_VALUE;
    }

    for (DataSetRow dataSetRow : dataSet.getRows()) {
        double[] input = dataSetRow.getInput();
        for (int i = 0; i < inputSize; i++) {
            if (input[i] > maxIn[i]) {
                maxIn[i] = input[i];
            }
            if (input[i] < minIn[i]) {
                minIn[i] = input[i];
            }
        }

        double[] output = dataSetRow.getDesiredOutput();
        for (int i = 0; i < outputSize; i++) {
            if (output[i] > maxOut[i]) {
                maxOut[i] = output[i];
            }
            if (output[i] < minOut[i]) {
                minOut[i] = output[i];
            }
        }

    }
}
 

/**
* Initialize normalizer: finds min and max values for all the columns in the data set.
* 
* @param dataSet 
*/
private void init(DataSet dataSet) {
    int numInputs = dataSet.getInputSize();
    int numOutputs = dataSet.getOutputSize();

    maxIn = new double[numInputs];
    minIn = new double[numInputs];

    for (int i = 0; i < numInputs; i++) {
        maxIn[i] = Double.MIN_VALUE;
        minIn[i] = Double.MAX_VALUE;
    }

    maxOut = new double[numOutputs];
    minOut = new double[numOutputs];

    for (int i = 0; i < numOutputs; i++) {
        maxOut[i] = Double.MIN_VALUE;
        minOut[i] = Double.MAX_VALUE;
    }

    for (DataSetRow dataSetRow : dataSet.getRows()) {
        double[] input = dataSetRow.getInput();
        for (int i = 0; i < numInputs; i++) {
            if (Math.abs(input[i]) > maxIn[i]) {
                maxIn[i] = Math.abs(input[i]);
            }
            if (Math.abs(input[i]) < minIn[i]) {
                minIn[i] = Math.abs(input[i]);
            }
        }

        double[] output = dataSetRow.getDesiredOutput();
        for (int i = 0; i < numOutputs; i++) {
            if (Math.abs(output[i]) > maxOut[i]) {
                maxOut[i] = Math.abs(output[i]);
            }
            if (Math.abs(output[i]) < minOut[i]) {
                minOut[i] = Math.abs(output[i]);
            }
        }
    }
}
 

/**
     *
     * You can get results calling getResults() method.
     *
     * @param neuralNetwork type of neural net
     * @param dataSet
     */
    public void train(DataSet dataSet) {// mozda da se vrati Training setting koji je najbolje resenje za dati dataset.??
        generateTrainingSettings();
        List<TrainingResult> statResults = null;
        DataSet trainingSet, testSet; // validationSet;

        if (splitTrainTest) {
            DataSet[] dataSplit = dataSet.split(splitPercentage, 100-splitPercentage); //opet ne radi Maven za neuroph 2.92
            trainingSet = dataSplit[0];
            testSet = dataSplit[1];
        } else {
            trainingSet = dataSet;
            testSet = dataSet;
        }

        if (generateStatistics) {
            statResults = new ArrayList<>();
        }

        int trainingNo = 0;
        for (TrainingSettings trainingSetting : trainingSettingsList) {
            System.out.println("-----------------------------------------------------------------------------------");
            trainingNo++;
            System.out.println("##TRAINING: " + trainingNo);
            trainingSetting.setTrainingSet(splitPercentage);
            trainingSetting.setTestSet(100 - splitPercentage);
            //int subtrainNo = 0;

            for (int subtrainNo = 1; subtrainNo <= repeat; subtrainNo++) {
                System.out.println("#SubTraining: " + subtrainNo);

                MultiLayerPerceptron neuralNet
                        = new MultiLayerPerceptron(dataSet.getInputSize(), trainingSetting.getHiddenNeurons(), dataSet.getOutputSize());

                BackPropagation bp = neuralNet.getLearningRule();

                bp.setLearningRate(trainingSetting.getLearningRate());
                bp.setMaxError(trainingSetting.getMaxError());
                bp.setMaxIterations(trainingSetting.getMaxIterations());

                neuralNet.learn(trainingSet);
//                  testNeuralNetwork(neuralNet, testSet); // not implemented
                ConfusionMatrix cm = new ConfusionMatrix(new String[]{""});
                TrainingResult result = new TrainingResult(trainingSetting, bp.getTotalNetworkError(), bp.getCurrentIteration(),cm);
                System.out.println(subtrainNo + ") iterations: " + bp.getCurrentIteration());

                if (generateStatistics) {
                    statResults.add(result);
                } else {
                    results.add(result);
                }

            }

            if (generateStatistics) {
                TrainingResult trainingStats = calculateTrainingStatistics(trainingSetting, statResults);
                results.add(trainingStats);
                statResults.clear();
            }

        }

    }