Created
June 30, 2016 11:48
-
-
Save fornarat/8748219796cf48c28c62f6f5851f617e to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
package org.deeplearning4j.examples.convolution; | |
import org.canova.api.records.reader.RecordReader; | |
import org.canova.api.records.reader.impl.CSVRecordReader; | |
import org.canova.api.split.FileSplit; | |
import org.deeplearning4j.datasets.canova.RecordReaderDataSetIterator; | |
import org.deeplearning4j.datasets.iterator.impl.MnistDataSetIterator; | |
import org.deeplearning4j.eval.Evaluation; | |
import org.deeplearning4j.nn.api.OptimizationAlgorithm; | |
import org.deeplearning4j.nn.conf.MultiLayerConfiguration; | |
import org.deeplearning4j.nn.conf.NeuralNetConfiguration; | |
import org.deeplearning4j.nn.conf.Updater; | |
import org.deeplearning4j.nn.conf.layers.ConvolutionLayer; | |
import org.deeplearning4j.nn.conf.layers.DenseLayer; | |
import org.deeplearning4j.nn.conf.layers.OutputLayer; | |
import org.deeplearning4j.nn.conf.layers.SubsamplingLayer; | |
import org.deeplearning4j.nn.conf.layers.setup.ConvolutionLayerSetup; | |
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork; | |
import org.deeplearning4j.nn.weights.WeightInit; | |
import org.deeplearning4j.optimize.listeners.ScoreIterationListener; | |
import org.nd4j.linalg.api.ndarray.INDArray; | |
import org.nd4j.linalg.dataset.DataSet; | |
import org.nd4j.linalg.dataset.api.iterator.DataSetIterator; | |
import org.nd4j.linalg.dataset.api.preprocessor.NormalizerMinMaxScaler; | |
import org.nd4j.linalg.lossfunctions.LossFunctions; | |
import org.slf4j.Logger; | |
import org.slf4j.LoggerFactory; | |
//import org.deeplearning4j.nn.conf.LearningRatePolicy; | |
import java.io.File; | |
/** | |
* Created by agibsonccc on 9/16/15. | |
* modified by fornarat on 06/30/16 | |
*/ | |
public class LenetMnistExampleCustom { | |
private static final Logger log = LoggerFactory.getLogger(LenetMnistExampleCustom.class); | |
public static void main(String[] args) throws Exception { | |
/* | |
int nChannels = 1; | |
int outputNum = 10; | |
int batchSize = 64; | |
int nEpochs = 10; | |
int iterations = 1; | |
int seed = 123; | |
*/ | |
int iterations = 1; | |
int nChannels = 1; | |
int seed = 123; | |
double learningRate = 0.01; | |
int batchSize = 3500; | |
int nEpochs = 30; | |
// int numInputs = 2; | |
int outputNum = 2; | |
// int numHiddenNodes = 20; | |
log.info("Load data...."); | |
RecordReader rrTrain = new CSVRecordReader(); | |
rrTrain.initialize(new FileSplit(new File("src/main/resources/classification/train.csv"))); | |
org.deeplearning4j.datasets.iterator.DataSetIterator dataSetIteratorTrain = new RecordReaderDataSetIterator(rrTrain,batchSize,0,2); | |
NormalizerMinMaxScaler preProcessor = new NormalizerMinMaxScaler(); | |
log.info("During 'fit' the preprocessor calculates the metrics (std dev and mean for the standardizer, min and max for minmaxscaler) from the data given"); | |
log.info("Fit can take a dataset or a dataset iterator\n"); | |
//Fitting a preprocessor with a dataset | |
log.info("Fitting with a dataset..............."); | |
// DataSet trainDataset = dataSetIteratorTrain.next(); | |
preProcessor.fit(dataSetIteratorTrain); | |
preProcessor.transform(dataSetIteratorTrain); | |
//Load the test/evaluation data: | |
RecordReader rrTest = new CSVRecordReader(); | |
rrTest.initialize(new FileSplit(new File("src/main/resources/classification/test.csv"))); | |
org.deeplearning4j.datasets.iterator.DataSetIterator dataSetIteratorTest = new RecordReaderDataSetIterator(rrTest,batchSize,0,2); | |
preProcessor.fit(dataSetIteratorTest); | |
preProcessor.transform(dataSetIteratorTest); | |
log.info("Build model...."); | |
MultiLayerConfiguration.Builder builder = new NeuralNetConfiguration.Builder() | |
.seed(seed) | |
.iterations(iterations) | |
.regularization(true).l2(0.0005) | |
.learningRate(learningRate)//.biasLearningRate(0.02) | |
//.learningRateDecayPolicy(LearningRatePolicy.Inverse).lrPolicyDecayRate(0.001).lrPolicyPower(0.75) | |
.weightInit(WeightInit.XAVIER) | |
.optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT) | |
.updater(Updater.NESTEROVS).momentum(0.9) | |
.list() | |
.layer(0, new ConvolutionLayer.Builder(5, 5) | |
.nIn(nChannels) | |
.stride(1, 1) | |
.nOut(20) | |
// .nOut(outputNum) | |
.activation("identity") | |
.build()) | |
.layer(1, new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX) | |
.kernelSize(2,2) | |
.stride(2,2) | |
.build()) | |
.layer(2, new ConvolutionLayer.Builder(5, 5) | |
.nIn(nChannels) | |
.stride(1, 1) | |
.nOut(50) | |
.activation("identity") | |
.build()) | |
.layer(3, new SubsamplingLayer.Builder(SubsamplingLayer.PoolingType.MAX) | |
.kernelSize(2,2) | |
.stride(2,2) | |
.build()) | |
.layer(4, new DenseLayer.Builder().activation("relu") | |
.nOut(500).build()) | |
.layer(5, new OutputLayer.Builder(LossFunctions.LossFunction.NEGATIVELOGLIKELIHOOD) | |
.nOut(outputNum) | |
.activation("softmax") | |
.build()) | |
.backprop(true).pretrain(false); | |
new ConvolutionLayerSetup(builder,100, 90,1); | |
MultiLayerConfiguration conf = builder.build(); | |
MultiLayerNetwork model = new MultiLayerNetwork(conf); | |
model.init(); | |
log.info("Train model...."); | |
model.setListeners(new ScoreIterationListener(1)); | |
for( int i=0; i<nEpochs; i++ ) { | |
model.fit(dataSetIteratorTrain); | |
log.info("*** Completed epoch {} ***", i); | |
log.info("Evaluate model...."); | |
Evaluation eval = new Evaluation(outputNum); | |
while(dataSetIteratorTest.hasNext()){ | |
DataSet ds = dataSetIteratorTest.next(); | |
INDArray output = model.output(ds.getFeatureMatrix(), false); | |
eval.eval(ds.getLabels(), output); | |
} | |
log.info(eval.stats()); | |
dataSetIteratorTest.reset(); | |
} | |
log.info("****************Example finished********************"); | |
} | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment