package org.nd4j.examples; import org.nd4j.linalg.api.ndarray.INDArray; import org.nd4j.linalg.factory.Nd4j; import org.nd4j.linalg.indexing.INDArrayIndex; import org.nd4j.linalg.indexing.NDArrayIndex; import java.util.Arrays; /** * --- Nd4j Example 3: Getting and setting parts of INDArrays --- * * In this example, we'll see ways to obtain and manipulate subsets of INDArray * * @author Alex Black */ public class Nd4jEx3_GettingAndSettingSubsets { public static void main(String[] args){ //Let's start by creating a 3x5 INDArray with manually specified values // To do this, we are starting with a 1x15 array, and perform a 'reshape' operation to convert it to a 3x5 INDArray INDArray originalArray = Nd4j.linspace(1,15,15).reshape('c',3,5); System.out.println("Original Array:"); System.out.println(originalArray); //We can use getRow and getColumn operations to get a row or column respectively: INDArray firstRow = originalArray.getRow(0); INDArray lastColumn = originalArray.getColumn(4); System.out.println(); System.out.println("First row:\n" + firstRow); System.out.println("Last column:\n" + lastColumn); //Careful of the printing here: lastColumn looks like a row vector when printed, but it's really a column vector System.out.println("Shapes: " + Arrays.toString(firstRow.shape()) + "\t" + Arrays.toString(lastColumn.shape())); //A key concept in ND4J is the idea of views: one INDArray may point to the same locations in memory as other arrays //For example, getRow and getColumn are both views of originalArray //Consequently, changes to one results in changes to the other: firstRow.addi(1.0); //In-place addition operation: changes the values of both firstRow AND originalArray: System.out.println("\n\n"); System.out.println("firstRow, after addi operation:"); System.out.println(firstRow); System.out.println("originalArray, after firstRow.addi(1.0) operation: (note it is modified, as firstRow is a view of originalArray)"); System.out.println(originalArray); //Let's recreate our our original array for the next section... originalArray = Nd4j.linspace(1,15,15).reshape('c',3,5); //We can select arbitrary subsets, using INDArray indexing: //All rows, first 3 columns (note that internal here is columns 0 inclusive to 3 exclusive) INDArray first3Columns = originalArray.get(NDArrayIndex.all(), NDArrayIndex.interval(0,3)); System.out.println("first 3 columns:\n" + first3Columns); //Again, this is also a view: first3Columns.addi(100); System.out.println("originalArray, after first3Columns.addi(100)"); System.out.println(originalArray); //Let's recreate our our original array for the next section... originalArray = Nd4j.linspace(1,15,15).reshape('c',3,5); //We can similarly set arbitrary subsets. //Let's set the 3rd column (index 2) to zeros: INDArray zerosColumn = Nd4j.zeros(3,1); originalArray.put(new INDArrayIndex[]{NDArrayIndex.all(), NDArrayIndex.point(2)}, zerosColumn); //All rows, column index 2 System.out.println("\n\n\nOriginal array, after put operation:\n" + originalArray); //Let's recreate our our original array for the next section... originalArray = Nd4j.linspace(1,15,15).reshape('c',3,5); //Sometimes, we don't want this in-place behaviour. In this case: just add a .dup() operation at the end //the .dup() operation - aka 'duplicate' - creates a new and separate array INDArray firstRowDup = originalArray.getRow(0).dup(); //We now have a copy of the first row. i.e., firstRowDup is NOT a view of originalArray firstRowDup.addi(100); System.out.println("\n\n\n"); System.out.println("firstRowDup, after .addi(100):\n" + firstRowDup); System.out.println("originalArray, after firstRowDup.addi(100): (note it is unmodified)\n" + originalArray); } }