Java example source code file (LocalResponseTest.java)

This example Java source code file (LocalResponseTest.java) is included in the alvinalexander.com "Java Source Code Warehouse" project. The intent of this project is to help you "Learn Java by Example" ^TM.

Learn more about this Java project at its project page.

Java - Java tags/keywords

before, exception, indarray, localresponsetest, mnistdatasetiterator, multilayerconfiguration, multilayernetwork, neuralnetconfiguration, pair, test

The LocalResponseTest.java Java example source code

package org.deeplearning4j.nn.layers.normalization;

import org.deeplearning4j.berkeley.Pair;
import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
import org.deeplearning4j.datasets.iterator.impl.MnistDataSetIterator;
import org.deeplearning4j.nn.api.Layer;
import org.deeplearning4j.nn.api.OptimizationAlgorithm;
import org.deeplearning4j.nn.conf.GradientNormalization;
import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.layers.ConvolutionLayer;
import org.deeplearning4j.nn.conf.layers.DenseLayer;
import org.deeplearning4j.nn.conf.layers.LocalResponseNormalization;
import org.deeplearning4j.nn.conf.layers.OutputLayer;
import org.deeplearning4j.nn.gradient.Gradient;
import org.deeplearning4j.nn.layers.factory.LayerFactories;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.nn.weights.WeightInit;
import org.junit.Before;
import org.junit.Test;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.dataset.DataSet;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.lossfunctions.LossFunctions;

import static org.junit.Assert.*;

/**
 *
 */
public class LocalResponseTest {

    private INDArray x = Nd4j.create(new double[]{
            0.88128096, -0.96666986, -0.61832994,  0.26418415,  0.05694608,
            0.2950289 ,  0.99222249,  0.24541704,  0.4219842 ,  0.96430975,
            0.19299535, -0.06658337, -0.27603117,  0.24216647,  0.21834095,
            0.03863283, -0.82313406, -0.37236378, -0.77667993,  0.66295379,
            -0.34406275, -0.25924176,  0.26652309, -0.58964926, -0.46907067,
            0.34666502,  0.81208313, -0.17042427, -0.22470538,  0.8348338 ,
            0.50494033,  0.45004508,  0.58735144, -0.87217808, -0.74788797,
            -0.04363599,  0.72276866,  0.52476895, -0.52383977,  0.1311436 ,
            0.2628099 ,  0.77274454,  0.86400729, -0.35246921, -0.03399619,
            -0.502312  ,  0.42834607,  0.85534132,  0.90083021,  0.24571614,
            0.63058525, -0.82919437,  0.57236177, -0.0913529 , -0.7102778 ,
            0.81631756, -0.89004314,  0.43995622, -0.26112801, -0.76135367,
            0.65180862, -0.54667377,  0.94908774,  0.59298772,  0.36457643,
            0.58892179, -0.52951556,  0.31559938, -0.55268252,  0.8272332 ,
            0.37911707, -0.96299696, -0.40717798,  0.43324658,  0.2589654 ,
            -0.15605508,  0.96334064, -0.31666604,  0.19781154,  0.09908111,
            0.64796048, -0.99037546,  0.67919868,  0.43810204
    }, new int[] {2,7,3,2});

    private INDArray activationsExpected = Nd4j.create(new double[]{
            0.52397668, -0.57476264, -0.3676528 ,  0.15707894,  0.03385943,
            0.17542371,  0.58992499,  0.14591768,  0.25090647,  0.57335907,
            0.11475233, -0.03958985, -0.16411273,  0.14398433,  0.12981956,
            0.02297027, -0.48942304, -0.22139823, -0.46177959,  0.39418164,
            -0.20457059, -0.15413573,  0.15846729, -0.3505919 , -0.27889356,
            0.20611978,  0.48284137, -0.10133155, -0.13360347,  0.49636194,
            0.30022132,  0.26758799,  0.34922296, -0.51858318, -0.4446843 ,
            -0.02594452,  0.42974478,  0.31202248, -0.31146204,  0.07797609,
            0.15626372,  0.4594543 ,  0.51370209, -0.20957276, -0.02021335,
            -0.29866382,  0.25469059,  0.50856382,  0.53558689,  0.14609739,
            0.37491882, -0.49301448,  0.34031925, -0.05431537, -0.42228988,
            0.48536259, -0.52917528,  0.26157826, -0.15526266, -0.45265958,
            0.38753596, -0.32503816,  0.56427884,  0.35256693,  0.21676543,
            0.35014921, -0.31483513,  0.18764766, -0.32859638,  0.49183461,
            0.22540972, -0.57255536, -0.24210122,  0.25760418,  0.15397197,
            -0.0927838 ,  0.57277   , -0.18827969,  0.1176173 ,  0.05891332,
            0.38526815, -0.58884346,  0.40383074,  0.26048511
    }, new int[] {2,7,3,2});

    private INDArray epsilon = Nd4j.create(new double[] {
            -0.13515499,  0.96470547, -0.62253004,  0.80172491, -0.97510445,
            -0.41198033, -0.4790071 ,  0.07551047, -0.01383764, -0.05797465,
            0.21242172,  0.7145375 , -0.17809176, -0.11465316, -0.2066526 ,
            0.21950938,  0.4627091 ,  0.30275798,  0.61443841,  0.75912178,
            -0.132248  , -0.82923287,  0.74962652, -0.88993639,  0.04406403,
            0.32096064, -0.46400586,  0.1603231 ,  0.63007826,  0.10626783,
            0.08009516,  0.88297033,  0.11441587,  0.35862735,  0.40441504,
            -0.60132015,  0.87743825,  0.09792926,  0.92742652,  0.6182847 ,
            -0.9602651 , -0.19611064,  0.15762019,  0.00339905, -0.9238292 ,
            0.02451134, -0.44294646, -0.5450229 ,  0.87502575, -0.59481794,
            0.65259099, -0.77772689,  0.53300053,  0.11541174,  0.32667685,
            0.99437004, -0.04084824, -0.45166185,  0.29513556,  0.53582036,
            0.95541358, -0.75714606, -0.63295805, -0.70315111, -0.6553846 ,
            -0.78824568,  0.84295344, -0.38352135, -0.04541624,  0.17396702,
            0.41530582,  0.11870354,  0.85787249, -0.94597596,  0.05792254,
            0.04811822,  0.04847952, -0.82953823,  0.8089835 ,  0.50185651,
            -0.88619858, -0.78598201,  0.27489874,  0.63673472
    }, new int[] {2,7,3,2});

    private INDArray newEpsilonExpected = Nd4j.create(new double[]{
            -0.08033668,  0.57355404, -0.37014094,  0.47668865, -0.57978398,
            -0.24495915, -0.28474802,  0.04490108, -0.00823483, -0.03448687,
            0.12630466,  0.42485803, -0.10589627, -0.06816553, -0.12287001,
            0.13051508,  0.27510744,  0.18001786,  0.36528736,  0.45133191,
            -0.07863599, -0.49303374,  0.44571424, -0.52912313,  0.02620371,
            0.19082049, -0.27585581,  0.09532529,  0.3746179 ,  0.06316902,
            0.04761803,  0.52497554,  0.06804816,  0.21323238,  0.24044329,
            -0.35752413,  0.52168733,  0.05821467,  0.55140609,  0.3676247 ,
            -0.57095432, -0.11660115,  0.09367896,  0.00202246, -0.54928631,
            0.01455687, -0.26336867, -0.3240425 ,  0.52023786, -0.35366109,
            0.3879728 , -0.46243483,  0.31692421,  0.06862034,  0.19421607,
            0.59124804, -0.0242459 , -0.26852599,  0.17547797,  0.31857637,
            0.56804365, -0.45020312, -0.37634474, -0.41804832, -0.38966343,
            -0.4686695 ,  0.50119156, -0.22802454, -0.02698562,  0.10343311,
            0.24693431,  0.0706142 ,  0.5100745 , -0.56245267,  0.03443092,
            0.02860913,  0.02883426, -0.49320197,  0.4810102 ,  0.29840365,
            -0.5269345 , -0.46732581,  0.16344811,  0.37857518
    }, new int[] {2,7,3,2});

    private INDArray activationsActual;
    private Layer layer;

    @Before
    public void doBefore(){
        NeuralNetConfiguration conf = new NeuralNetConfiguration.Builder()
                .gradientNormalization(GradientNormalization.RenormalizeL2PerLayer)
                .seed(123)
                .layer(new LocalResponseNormalization.Builder()
                        .k(2).n(5).alpha(1e-4).beta(0.75)
                        .build())
                .build();

        layer = LayerFactories.getFactory(new LocalResponseNormalization()).create(conf, null, 0, null, false);
        activationsActual = layer.activate(x);
    }

    @Test
    public void testActivate(){
        // Precision is off from the expected results because expected results generated in numpy
        assertEquals(activationsExpected, activationsActual);
        assertArrayEquals(activationsExpected.shape(), activationsActual.shape());
        }

    @Test
    public void testBackpropGradient(){
        Pair<Gradient, INDArray> containedOutput = layer.backpropGradient(epsilon);

        assertEquals(newEpsilonExpected.getDouble(8), containedOutput.getSecond().getDouble(8), 1e-4);
        assertEquals(newEpsilonExpected.getDouble(20), containedOutput.getSecond().getDouble(20), 1e-4);
        assertEquals(null, containedOutput.getFirst().getGradientFor("W"));
        assertArrayEquals(newEpsilonExpected.shape(), containedOutput.getSecond().shape());
    }

    @Test
    public void testRegularization(){
        // Confirm a structure with regularization true will not throw an error

        NeuralNetConfiguration conf = new NeuralNetConfiguration.Builder()
                .gradientNormalization(GradientNormalization.RenormalizeL2PerLayer)
                .regularization(true)
                .l1(0.2)
                .l2(0.1)
                .seed(123)
                .layer(new LocalResponseNormalization.Builder()
                        .k(2).n(5).alpha(1e-4).beta(0.75)
                        .build())
                .build();
    }

    @Test
    public void testMultiCNNLayer() throws Exception {
        MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
                .optimizationAlgo(OptimizationAlgorithm.LINE_GRADIENT_DESCENT)
                .iterations(1)
                .seed(123)
                .list()
                .layer(0, new ConvolutionLayer.Builder().nIn(1).nOut(6).weightInit(WeightInit.XAVIER).activation("relu").build())
                .layer(1, new LocalResponseNormalization.Builder().build())
                .layer(2, new DenseLayer.Builder().nOut(2).build())
                .layer(3, new OutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
                        .weightInit(WeightInit.XAVIER)
                        .activation("softmax")
                        .nIn(2).nOut(10).build())
                .backprop(true).pretrain(false)
                .cnnInputSize(28,28,1)
                .build();

        MultiLayerNetwork network = new MultiLayerNetwork(conf);
        network.init();
        DataSetIterator iter = new MnistDataSetIterator(2, 2);
        DataSet next = iter.next();

        network.fit(next);


    }


}