Field Detail

• vals

  protected final float[][] vals

  The dataset, first indexed by attribute, then by instance.

• attNumVals

  protected final int[] attNumVals

  Attribute description - holds a 0 for numeric attributes, and the number of available categories for nominal attributes.

• classIndex

  protected final int classIndex

  Numeric index of the class attribute.

• numAttributes

  protected final int numAttributes

  Number of attributes, including the class attribute.

• numClasses

  protected final int numClasses

  Number of classes.

• numInstances

  protected final int numInstances

  Number of instances.

• instClassValues

  protected final int[] instClassValues

  The class an instance belongs to.

• sortedIndices

  protected int[][] sortedIndices

  Ordering of instances, indexed by attribute, then by instance.

• instWeights

  protected double[] instWeights

  Weights of instances.

• inBag

  protected boolean[] inBag

  Is instance in 'bag' created by bootstrap sampling.

• numInBag

  protected int numInBag

  How many instances are in 'bag' created by bootstrap sampling.

• whatGoesWhere

  protected int[] whatGoesWhere

  Used in training of FastRandomTrees.

• reusableRandomGenerator

  protected Random reusableRandomGenerator

  Used in training of FastRandomTrees. Each tree can store its own custom-seeded random generator in this field.

Constructor Detail

• DataCache

  public DataCache(Instances origData)
            throws Exception

  Creates a DataCache by copying data from a weka.core.Instances object.

  Throws:

  Exception

• DataCache

  public DataCache(DataCache origData)

  Makes a copy of a DataCache. Most array fields are shallow copied, with the exception of in inBag and whatGoesWhere arrays, which are created anew.

  Parameters:

  origData -

Method Detail

• scrambleOneAttribute

  public float[] scrambleOneAttribute(int attIndex,
                                      Random random)

  Randomizes one attribute in the vals[][]; returns a copy of the vals[] before randomization.

• resample

  public DataCache resample(int bagSize,
                            Random random)

  Uses sampling with replacement to create a new DataCache from an existing one. The probability of sampling a specific instance does not depend on its weight. When an instance is sampled multiple times, its weight in the new DataCache increases to a multiple of the original weight.

  Parameters:

  bagSize - If this is equal to the DataCache.numInstances, makes a a bootstrap sample (n of of in

  random - A random number generator.

  Returns:

  a new DataCache - consult "DataCache(DataCache origData)" constructor to see what's deep / shallow copied

• createInBagSortedIndices

  protected void createInBagSortedIndices()

  Invoked only when tree is trained.

• isValueMissing

  public final boolean isValueMissing(int attIndex,
                                      int instIndex)

  Does the given attribute - instance combination contain a missing value?

• isAttrNominal

  public final boolean isAttrNominal(int attIndex)

  Is an attribute with the given index nominal?

• getRandomNumberGenerator

  public Random getRandomNumberGenerator(long seed)

  Returns a random number generator. The initial seed of the random number generator depends on the given seed and the contents of the sortedIndices array (a single attribute is picked, its sortedIndices converted to String and a hashcode computed).

  Parameters:

  seed - the given seed

  Returns:

  the random number generator