In this tutorial we will use a neural network to forecast daily sea temperatures. The data consists of 2-dimensional temperature grids of 8 seas: Bengal, Korean, Black, Mediterranean, Arabian, Japan, Bohai, and Okhotsk Seas from 1981 to 2017. The raw data was taken from the Earth System Research Laboratory (https://www.esrl.noaa.gov/psd/) and preprocessed into CSV file. Each example consists of fifty 2-dimensional temperature grids, and every grid is represented by a single row in a CSV file. Thus, each sequence is represented by a CSV file with 50 rows.
For this task, we will use a convolutional LSTM neural network to forecast next-day sea temperatures for a given sequence of temperature grids. Recall, a convolutional network is most often used for image data like the MNIST dataset (dataset of handwritten images). A convolutional network is appropriate for this type of gridded data, since each point in the 2-dimensional grid is related to its neighbor points. Furthermore, the data is sequential, and each temperature grid is related to the previous grids. Because of these long and short term dependencies, a LSTM is fitting for this task too. For these two reasons, we will combine the aspects from these two different neural network architectures into a single convolutional LSTM network.
To download the data, we will create a temporary directory that will store the data files, extract the tar.gz file from the url, and place it in the specified directory.
val DATA_URL ="https://dl4jdata.blob.core.windows.net/training/seatemp/sea_temp.tar.gz"val DATA_PATH = FilenameUtils.concat(System.getProperty("java.io.tmpdir"), "dl4j_seas/")
We will then extract the data from the tar.gz file, recreate directories within the tar.gz file into our temporary directories, and copy the files from the tar.gz file.
var fileCount =0var dirCount =0val BUFFER_SIZE =4096val tais = new TarArchiveInputStream(new GzipCompressorInputStream( new BufferedInputStream( new FileInputStream(archizePath))))
var entry = tais.getNextEntry().asInstanceOf[TarArchiveEntry]while(entry !=null){if (entry.isDirectory()) {new File(DATA_PATH + entry.getName()).mkdirs() dirCount = dirCount +1 fileCount =0 }else {val data =new Array[scala.Byte](4* BUFFER_SIZE)val fos =new FileOutputStream(DATA_PATH + entry.getName());val dest =new BufferedOutputStream(fos, BUFFER_SIZE);var count = tais.read(data, 0, BUFFER_SIZE)while (count !=-1) { dest.write(data, 0, count) count = tais.read(data, 0, BUFFER_SIZE) } dest.close() fileCount = fileCount +1 }if(fileCount %1000==0){ print(".") } entry = tais.getNextEntry().asInstanceOf[TarArchiveEntry]}
DataSetIterators
Next we will convert the raw data (csv files) into DataSetIterators, which will be fed into a neural network. Our training data will have 1700 examples which will be represented by a single DataSetIterator, and the testing data will have 404 examples which will be represented by a separate DataSet Iterator.
val path = FilenameUtils.concat(DATA_PATH, "sea_temp/") // set parent directoryval featureBaseDir = FilenameUtils.concat(path, "features") // set feature directoryval targetsBaseDir = FilenameUtils.concat(path, "targets") // set label directory
We first initialize CSVSequenceRecordReaders, which will parse the raw data into record-like format. Then the SequenceRecordReaderDataSetIterators can be created using the RecordReaders. Since each example has exaclty 50 timesteps, an alignment mode of equal length is needed. Note also that this is a regression- based task and not a classification one.
The next task is to initialize the parameters for the convolutional LSTM neural network and then set up the neural network configuration.
val V_HEIGHT =13;val V_WIDTH =4;val kernelSize =2;val numChannels =1;
In the neural network configuraiton we will use the convolutional layer, LSTM layer, and output layer in success. In order to do this, we need to use the RnnToCnnPreProcessor and CnnToRnnPreprocessor. The RnnToCnnPreProcessor is used to reshape the 3-dimensional input from [batch size, height x width of grid, time series length ] into a 4 dimensional shape [number of examples x time series length , channels, width, height] which is suitable as input to a convolutional layer. The CnnToRnnPreProcessor is then used in a later layer to convert this convolutional shape back to the original 3-dimensional shape.
val conf =new NeuralNetConfiguration.Builder() .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT) .seed(12345) .weightInit(WeightInit.XAVIER) .updater(new AdaGrad(0.005)) .list() .layer(0, new ConvolutionLayer.Builder(kernelSize, kernelSize) .nIn(1) //1 channel .nOut(7) .stride(2, 2) .activation(Activation.RELU) .build()) .layer(1, new LSTM.Builder() .activation(Activation.SOFTSIGN) .nIn(84) .nOut(200) .gradientNormalization(GradientNormalization.ClipElementWiseAbsoluteValue) .gradientNormalizationThreshold(10) .build()) .layer(2, new RnnOutputLayer.Builder(LossFunction.MSE) .activation(Activation.IDENTITY) .nIn(200) .nOut(52) .gradientNormalization(GradientNormalization.ClipElementWiseAbsoluteValue) .gradientNormalizationThreshold(10) .build()) .inputPreProcessor(0, new RnnToCnnPreProcessor(V_HEIGHT, V_WIDTH, numChannels)) .inputPreProcessor(1, new CnnToRnnPreProcessor(6, 2, 7 )) .build();val net =new MultiLayerNetwork(conf);net.init();
Model Training
To train the model, we use 25 epochs and simply call the fit method of the MultiLayerNetwork.
// Train model on training setnet.fit(train , 25);
Model Evaluation
We will now evaluate our trained model. Note that we will use RegressionEvaluation, since our task is a regression and not a classification task.
val eval = net.evaluateRegression[RegressionEvaluation](test);test.reset();println()println( eval.stats() );
