Java example source code file (GravesLSTMOutputTest.java)

This example Java source code file (GravesLSTMOutputTest.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

afterclass, beforeclass, evaluation, graveslstmoutputtest, indarray, logger, multilayerconfiguration, multilayernetwork, negativedefaultstepfunction, random, rnntofeedforwardpreprocessor, scoreiterationlistener, test, type, util

The GravesLSTMOutputTest.java Java example source code

package org.deeplearning4j.nn.multilayer;

import org.deeplearning4j.eval.Evaluation;
import org.deeplearning4j.nn.api.OptimizationAlgorithm;
import org.deeplearning4j.nn.conf.BackpropType;
import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.Updater;
import org.deeplearning4j.nn.conf.distribution.NormalDistribution;
import org.deeplearning4j.nn.conf.layers.GravesLSTM;
import org.deeplearning4j.nn.conf.layers.OutputLayer;
import org.deeplearning4j.nn.conf.preprocessor.RnnToFeedForwardPreProcessor;
import org.deeplearning4j.nn.conf.stepfunctions.NegativeDefaultStepFunction;
import org.deeplearning4j.nn.weights.WeightInit;
import org.deeplearning4j.optimize.listeners.ScoreIterationListener;
import org.junit.AfterClass;
import org.junit.Assert;
import org.junit.BeforeClass;
import org.junit.Test;
import org.nd4j.linalg.api.buffer.DataBuffer.Type;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.indexing.NDArrayIndex;
import org.nd4j.linalg.lossfunctions.LossFunctions;
import org.nd4j.linalg.util.FeatureUtil;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.util.Random;

/**
 * Created by Kirill Lebedev (drlebedev.com) on 8/31/2015.
 */
public class GravesLSTMOutputTest {

    private static int nIn = 20;
    private static int layerSize = 15;
    private static int window = 300;
    private static INDArray data;
    private static Logger log;
    private static Type type;

    @BeforeClass
    public static void setUp() {
        type = Nd4j.dtype;
        Nd4j.dtype = Type.FLOAT;
        log = LoggerFactory.getLogger(GravesLSTMOutputTest.class);
        data = getData();
    }

    @AfterClass
    public static void tearDown() {
        data = null;
        log = null;
        Nd4j.dtype = type;
    }

    @Test
    public void testSameLabelsOutput() {
        MultiLayerNetwork network = new MultiLayerNetwork(getNetworkConf(40, false));
        network.init();
        network.setListeners(new ScoreIterationListener(1));
        network.fit(reshapeInput(data.dup()), data.dup());
        Evaluation ev = eval(network);
        Assert.assertTrue(ev.f1() > 0.90);
    }

    @Test
    public void testSameLabelsOutputWithTBPTT() {
        MultiLayerNetwork network = new MultiLayerNetwork(getNetworkConf(40, true));
        network.init();
        network.setListeners(new ScoreIterationListener(1));
        for (int i = 0; i < window / 100; i++) {
            INDArray d = data.get(NDArrayIndex.interval(100 * i, 100 * (i+1)), NDArrayIndex.all());
            network.fit(reshapeInput(d.dup()), reshapeInput(d.dup()));
        }
        Evaluation ev = eval(network);
    }

    private Evaluation eval(MultiLayerNetwork network) {
        Evaluation ev = new Evaluation(nIn);
        INDArray predict = network.output(reshapeInput(data));
        ev.eval(data, predict);
        log.info(ev.stats());
        return ev;
    }

    private MultiLayerConfiguration getNetworkConf(int iterations, boolean useTBPTT) {
        MultiLayerConfiguration.Builder builder = new NeuralNetConfiguration.Builder()
                .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)
                .learningRate(0.1)
                .regularization(true)
                .l2(0.0025)
                .iterations(iterations)
                .stepFunction(new NegativeDefaultStepFunction())
                .list()
                .layer(0, new GravesLSTM.Builder().weightInit(WeightInit.DISTRIBUTION)
                        .dist(new NormalDistribution(0.0, 0.01)).nIn(nIn).nOut(layerSize)
                        .updater(Updater.ADAGRAD)
                        .activation("tanh").build())
                .layer(1, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD)
                        .updater(Updater.ADAGRAD).nIn(layerSize).nOut(nIn)
                        .activation("softmax").build())
                .inputPreProcessor(1, new RnnToFeedForwardPreProcessor())
                .backprop(true)
                .pretrain(false);
        if (useTBPTT) {
            builder.backpropType(BackpropType.TruncatedBPTT);
            builder.tBPTTBackwardLength(window / 3);
            builder.tBPTTForwardLength(window / 3);
        }
        return builder.build();
    }

    private static INDArray getData() {
        Random r = new Random(1);
        int[] result = new int[window];
        for (int i = 0; i < window; i++) {
            result[i] = r.nextInt(nIn);
        }
        return FeatureUtil.toOutcomeMatrix(result, nIn);
    }

    private INDArray reshapeInput(INDArray inp) {
        int[] shape = inp.shape();
        int miniBatchSize = 1;
        INDArray reshaped = inp.reshape(miniBatchSize, shape[0] / miniBatchSize, shape[1]);
        return reshaped.permute(0, 2, 1);
    }
}