package com.intel.daal.examples.neural_networks;
import com.intel.daal.algorithms.neural_networks.layers.lcn.*;
import com.intel.daal.algorithms.neural_networks.layers.ForwardResultId;
import com.intel.daal.algorithms.neural_networks.layers.ForwardResultLayerDataId;
import com.intel.daal.algorithms.neural_networks.layers.ForwardInputId;
import com.intel.daal.algorithms.neural_networks.layers.BackwardResultId;
import com.intel.daal.algorithms.neural_networks.layers.BackwardInputId;
import com.intel.daal.algorithms.neural_networks.layers.BackwardInputLayerDataId;
import com.intel.daal.data_management.data.Tensor;
import com.intel.daal.data_management.data.HomogenTensor;
import com.intel.daal.examples.utils.Service;
import com.intel.daal.services.DaalContext;
class LCNLayerDenseBatch {
private static DaalContext context = new DaalContext();
public static void main(String[] args) throws java.io.FileNotFoundException, java.io.IOException {
long[] dimensionSizes = new long[4];
dimensionSizes[0] = 2;
dimensionSizes[1] = 1;
dimensionSizes[2] = 3;
dimensionSizes[3] = 4;
double[] data = new double[24];
Tensor dataTensor = new HomogenTensor(context, dimensionSizes, data, 1.0);
LcnForwardBatch lcnLayerForward = new LcnForwardBatch(context, Float.class, LcnMethod.defaultDense);
lcnLayerForward.input.set(ForwardInputId.data, dataTensor);
LcnForwardResult forwardResult = lcnLayerForward.compute();
Service.printTensor("Forward local contrast normalization layer result (first 5 rows):", forwardResult.get(ForwardResultId.value), 5, 0);
Service.printTensor("Centered data tensor (first 5 rows):", forwardResult.get(LcnLayerDataId.auxCenteredData), 5, 0);
Service.printTensor("Sigma tensor (first 5 rows):", forwardResult.get(LcnLayerDataId.auxSigma), 5, 0);
Service.printTensor("C tensor (first 5 rows):", forwardResult.get(LcnLayerDataId.auxC), 5, 0);
Service.printTensor("kernel:", lcnLayerForward.parameter.getKernel(), 5, 0);
Service.printNumericTable("getSumDimension:", lcnLayerForward.parameter.getSumDimension());
double[] backData = new double[24];
Tensor tensorDataBack = new HomogenTensor(context, dimensionSizes, backData, 0.01);
LcnBackwardBatch lcnLayerBackward = new LcnBackwardBatch(context, Float.class, LcnMethod.defaultDense);
lcnLayerBackward.input.set(BackwardInputId.inputGradient, tensorDataBack);
lcnLayerBackward.input.set(BackwardInputLayerDataId.inputFromForward,
forwardResult.get(ForwardResultLayerDataId.resultForBackward));
LcnBackwardResult backwardResult = lcnLayerBackward.compute();
Service.printTensor("Backward local contrast normalization layer result (first 5 rows):", backwardResult.get(BackwardResultId.gradient), 5, 0);
context.dispose();
}
}