hex
Class FrameTask<T extends FrameTask<T>>

java.lang.Object
  jsr166y.ForkJoinTask<java.lang.Void>
      jsr166y.CountedCompleter
          water.H2O.H2OCountedCompleter
              water.DTask
                  water.MRTask2<T>
                      hex.FrameTask<T>

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, java.util.concurrent.Future<java.lang.Void>, Freezable

Direct Known Subclasses:

GLMTask, Gram.GramTask

public abstract class FrameTask<T extends FrameTask<T>>
extends MRTask2<T>

See Also:

Serialized Form

Field Summary

protected int[]	_catOffsets
protected int	_cats
protected int	_nums
protected boolean	_standardize

Fields inherited from class water.MRTask2

_fr, _fs, _hi, _left, _lo, _nleft, _nodes, _nrite, _outputFrame, _rite, _topLocal

Fields inherited from class water.DTask

_cls, _eFromNode, _exception, _fname, _lineNum, _msg, _mth

Constructor Summary

protected	FrameTask(FrameTask ft)
	FrameTask(Job job, boolean standardize, boolean hasResponse)
	FrameTask(Job job, boolean standardize, boolean hasResponse, int step, int offset, boolean complement)

Method Summary

Frame	adaptFrame(Frame fr) Reorder the frame's columns so that numeric columns come first followed by categoricals ordered by cardinality in decreasing order and the response is the last.
int[]	catOffsets()
int	cats()
protected void	chunkDone() Override this to do post-chunk processing work.
protected void	chunkInit() Override this to initialize at the beginning of chunk processing.
T	dfork2(Frame fr)
T	doIt(Frame fr)
protected int	fullN()
protected int	largestCat()
void	map(Chunk[] chunks) Extracts the values, applies regularization to numerics, adds appropriate offsets to categoricals, and adapts response according to the CaseMode/CaseValue if set.
double[]	normMul()
double[]	normSub()
int	nums()
protected void	processRow(double[] nums, int ncats, int[] cats)
protected void	processRow(double[] nums, int ncats, int[] cats, double response) Method to process one row of the data for GLM functions.
boolean	standardize()

Methods inherited from class water.MRTask2

clone, closeLocal, compute2, dfork, dfork, dfork, dfork, dinvoke, doAll, doAll, doAll, doAll, getResult, map, map, map, map, map, map, map, map, map, map, map, onCompletion, onExceptionalCompletion, postGlobal, profString, reduce, reduce4, setupLocal, vecs

Methods inherited from class water.DTask

copyOver, frozenType, getDException, hasException, logVerbose, newInstance, onAck, onAckAck, read, setException, toDocField, write, writeJSONFields

Methods inherited from class water.H2O.H2OCountedCompleter

compute, priority

Methods inherited from class jsr166y.CountedCompleter

addToPendingCount, compareAndSetPendingCount, complete, exec, getCompleter, getPendingCount, getRawResult, setCompleter, setPendingCount, setRawResult, tryComplete

Methods inherited from class jsr166y.ForkJoinTask

adapt, adapt, adapt, cancel, compareAndSetForkJoinTaskTag, completeExceptionally, fork, get, get, getException, getForkJoinTaskTag, getPool, getQueuedTaskCount, getSurplusQueuedTaskCount, helpQuiesce, inForkJoinPool, invoke, invokeAll, invokeAll, invokeAll, isCancelled, isCompletedAbnormally, isCompletedNormally, isDone, join, peekNextLocalTask, pollNextLocalTask, pollTask, quietlyComplete, quietlyInvoke, quietlyJoin, reinitialize, setForkJoinTaskTag, tryUnfork

Methods inherited from class java.lang.Object

equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

_standardize

protected final boolean _standardize

_nums

protected int _nums

_cats

protected int _cats

_catOffsets

protected int[] _catOffsets

Constructor Detail

FrameTask

public FrameTask(Job job,
                 boolean standardize,
                 boolean hasResponse)

FrameTask

public FrameTask(Job job,
                 boolean standardize,
                 boolean hasResponse,
                 int step,
                 int offset,
                 boolean complement)

FrameTask

protected FrameTask(FrameTask ft)

Method Detail

standardize

public final boolean standardize()

nums

public final int nums()

cats

public final int cats()

catOffsets

public final int[] catOffsets()

normSub

public final double[] normSub()

normMul

public final double[] normMul()

fullN

protected final int fullN()

largestCat

protected final int largestCat()

processRow

protected void processRow(double[] nums,
                          int ncats,
                          int[] cats,
                          double response)

Method to process one row of the data for GLM functions. Numeric and categorical values are passed separately, as is reponse. Categoricals are passed as absolute indexes into the expanded beta vector, 0-levels are skipped (so the number of passed categoricals will not be the same for every row). Categorical expansion/indexing: Categoricals are placed in the beginning of the beta vector. Each cat variable with n levels is expanded into n-1 independent binary variables. Indexes in cats[] will point to the appropriate coefficient in the beta vector, so e.g. assume we have 2 categorical columns both with values A,B,C, then the following rows will have following indexes: A,A - ncats = 0, we do not pass any categorical here A,B - ncats = 1, indexes = [2] B,B - ncats = 2, indexes = [0,2] and so on

Parameters:

nums - - numeric values of this row

ncats - - number of passed (non-zero) categoricals

cats - - indexes of categoricals into the expanded beta-vector.

response - - numeric value for the response

processRow

protected void processRow(double[] nums,
                          int ncats,
                          int[] cats)

adaptFrame

public Frame adaptFrame(Frame fr)

Reorder the frame's columns so that numeric columns come first followed by categoricals ordered by cardinality in decreasing order and the response is the last.

Parameters:

fr -

Returns:

doIt

public T doIt(Frame fr)

dfork2

public T dfork2(Frame fr)

chunkInit

protected void chunkInit()

Override this to initialize at the beginning of chunk processing.

chunkDone

protected void chunkDone()

Override this to do post-chunk processing work.

map

public final void map(Chunk[] chunks)

Extracts the values, applies regularization to numerics, adds appropriate offsets to categoricals, and adapts response according to the CaseMode/CaseValue if set.

Overrides:

map in class MRTask2<T extends FrameTask<T>>