hex.tree.xgboost

Class XGBoostUtils

java.lang.Object

hex.tree.xgboost.XGBoostUtils

public class XGBoostUtils
extends java.lang.Object

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected static class	XGBoostUtils.SparseMatrix Sparse Matrix representation for XGBoost
protected static class	XGBoostUtils.SparseMatrixDimensions Dimensions of a Sparse Matrix

Field Summary

Fields

Modifier and Type	Field and Description
protected static int	SPARSE_MATRIX_DIM

Constructor Summary

Constructors

Constructor and Description
XGBoostUtils()

Method Summary

Modifier and Type	Method and Description
protected static XGBoostUtils.SparseMatrix	allocateCSRMatrix(XGBoostUtils.SparseMatrixDimensions sparseMatrixDimensions) Creates a XGBoostUtils.SparseMatrix object with pre-instantiated backing arrays for row-oriented compression schema (CSR).
static hex.tree.xgboost.XGBoostUtils.FeatureProperties	assembleFeatureNames(hex.DataInfo di)
protected static XGBoostUtils.SparseMatrixDimensions	calculateCSRMatrixDimensions(water.fvec.Chunk[] chunks, hex.DataInfo di, int weightColIndex)
protected static XGBoostUtils.SparseMatrixDimensions	calculateCSRMatrixDimensions(water.fvec.Frame f, int[] chunks, water.fvec.Vec.Reader[] vecs, water.fvec.Vec.Reader w, hex.DataInfo di)
static ml.dmlc.xgboost4j.java.DMatrix	convertChunksToDMatrix(hex.DataInfo di, water.fvec.Chunk[] chunks, int response, int weight, int fold, boolean sparse) convert a set of H2O chunks (representing a part of a vector) to a sparse DMatrix
static ml.dmlc.xgboost4j.java.DMatrix	convertFrameToDMatrix(hex.DataInfo di, water.fvec.Frame f, boolean onlyLocal, java.lang.String response, java.lang.String weight, java.lang.String fold, boolean sparse) convert an H2O Frame to a sparse DMatrix
protected static ml.dmlc.xgboost4j.java.DMatrix	csr(water.fvec.Chunk[] chunks, int weight, int respIdx, water.fvec.Frame f, int[] chunksIds, water.fvec.Vec.Reader[] vecs, water.fvec.Vec.Reader w, water.fvec.Vec.Reader respReader, int nRows, hex.DataInfo di, float[] resp, float[] weights)
protected static int	initalizeFromChunkIds(water.fvec.Frame f, int[] chunks, water.fvec.Vec.Reader[] vecs, water.fvec.Vec.Reader w, hex.DataInfo di, int actualRows, long[][] rowHeaders, float[][] data, int[][] colIndex, water.fvec.Vec.Reader respVec, float[] resp, float[] weights)
static java.lang.String	makeFeatureMap(water.fvec.Frame f, hex.DataInfo di)

Methods inherited from class java.lang.Object

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

Field Detail

SPARSE_MATRIX_DIM

protected static int SPARSE_MATRIX_DIM

Constructor Detail

XGBoostUtils

public XGBoostUtils()

Method Detail

makeFeatureMap

public static java.lang.String makeFeatureMap(water.fvec.Frame f,
                                              hex.DataInfo di)

convertFrameToDMatrix

public static ml.dmlc.xgboost4j.java.DMatrix convertFrameToDMatrix(hex.DataInfo di,
                                                                   water.fvec.Frame f,
                                                                   boolean onlyLocal,
                                                                   java.lang.String response,
                                                                   java.lang.String weight,
                                                                   java.lang.String fold,
                                                                   boolean sparse)
                                                            throws ml.dmlc.xgboost4j.java.XGBoostError

convert an H2O Frame to a sparse DMatrix

Parameters:

f - H2O Frame

onlyLocal - if true uses only chunks local to this node

response - name of the response column

weight - name of the weight column

fold - name of the fold assignment column

Returns:

DMatrix

Throws:

ml.dmlc.xgboost4j.java.XGBoostError

convertChunksToDMatrix

public static ml.dmlc.xgboost4j.java.DMatrix convertChunksToDMatrix(hex.DataInfo di,
                                                                    water.fvec.Chunk[] chunks,
                                                                    int response,
                                                                    int weight,
                                                                    int fold,
                                                                    boolean sparse)
                                                             throws ml.dmlc.xgboost4j.java.XGBoostError

convert a set of H2O chunks (representing a part of a vector) to a sparse DMatrix

Parameters:

response - name of the response column

weight - name of the weight column

fold - name of the fold assignment column

Returns:

DMatrix

Throws:

ml.dmlc.xgboost4j.java.XGBoostError

csr

protected static ml.dmlc.xgboost4j.java.DMatrix csr(water.fvec.Chunk[] chunks,
                                                    int weight,
                                                    int respIdx,
                                                    water.fvec.Frame f,
                                                    int[] chunksIds,
                                                    water.fvec.Vec.Reader[] vecs,
                                                    water.fvec.Vec.Reader w,
                                                    water.fvec.Vec.Reader respReader,
                                                    int nRows,
                                                    hex.DataInfo di,
                                                    float[] resp,
                                                    float[] weights)
                                             throws ml.dmlc.xgboost4j.java.XGBoostError

Throws:

ml.dmlc.xgboost4j.java.XGBoostError

initalizeFromChunkIds

protected static int initalizeFromChunkIds(water.fvec.Frame f,
                                           int[] chunks,
                                           water.fvec.Vec.Reader[] vecs,
                                           water.fvec.Vec.Reader w,
                                           hex.DataInfo di,
                                           int actualRows,
                                           long[][] rowHeaders,
                                           float[][] data,
                                           int[][] colIndex,
                                           water.fvec.Vec.Reader respVec,
                                           float[] resp,
                                           float[] weights)

allocateCSRMatrix

protected static XGBoostUtils.SparseMatrix allocateCSRMatrix(XGBoostUtils.SparseMatrixDimensions sparseMatrixDimensions)

Creates a XGBoostUtils.SparseMatrix object with pre-instantiated backing arrays for row-oriented compression schema (CSR). All backing arrays are allocated using MemoryManager.

Parameters:

sparseMatrixDimensions - Dimensions of a sparse matrix

Returns:

An instance of XGBoostUtils.SparseMatrix with pre-allocated backing arrays.

calculateCSRMatrixDimensions

protected static XGBoostUtils.SparseMatrixDimensions calculateCSRMatrixDimensions(water.fvec.Chunk[] chunks,
                                                                                  hex.DataInfo di,
                                                                                  int weightColIndex)

calculateCSRMatrixDimensions

protected static XGBoostUtils.SparseMatrixDimensions calculateCSRMatrixDimensions(water.fvec.Frame f,
                                                                                  int[] chunks,
                                                                                  water.fvec.Vec.Reader[] vecs,
                                                                                  water.fvec.Vec.Reader w,
                                                                                  hex.DataInfo di)

assembleFeatureNames

public static hex.tree.xgboost.XGBoostUtils.FeatureProperties assembleFeatureNames(hex.DataInfo di)