org.knime.base.node.mine.bfn
Class BasisFunctionFactory

java.lang.Object
  org.knime.base.node.mine.bfn.BasisFunctionFactory

Direct Known Subclasses:

FuzzyBasisFunctionFactory, RadialBasisFunctionFactory

public abstract class BasisFunctionFactory
extends Object

Factory class for BasisFunctionLearnerRow which automatically creates new basis functions of a certain type.

Author:

Thomas Gabriel, University of Konstanz

See Also:

BasisFunctionLearnerRow

Field Summary

(package private) static String	CFG_DISTANCE Key for the distance function.
(package private) static String	CFG_MODEL_SPEC Key for the model spec.
static DataColumnSpec	CLASS_COLUMN Name of the basisfunction class column.

Constructor Summary

protected

BasisFunctionFactory(DataTableSpec spec, String[] targetColumns, DataType type, int distance) Creates new basisfunction factory with the given spec to extract min/max value for all numeric columns.

Method Summary

abstract BasisFunctionLearnerRow	commit(RowKey key, DataCell classInfo, DataRow row) Returns a new row initialised by a DataRow as its initial center vector and a class label.
static DataTableSpec	createModelSpec(DataTableSpec inSpec, String[] dataColumns, String[] targetColumns, DataType type) Creates a model spec based on the data input spec by extracting all DoubleCell columns and the specified target column.
static String[]	findDataColumns(DataTableSpec spec, List<String> targetCols) Find all numeric columns which are not target columns.
int	getDistance() Returns the choice of distance function.
MutableDouble[]	getMaximums()
MutableDouble[]	getMinimums()
DataTableSpec	getModelSpec()
void	save(ModelContent pp) Saves to model content.

Methods inherited from class java.lang.Object

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

Field Detail

CFG_DISTANCE

static final String CFG_DISTANCE

Key for the distance function.

See Also:

Constant Field Values

CFG_MODEL_SPEC

static final String CFG_MODEL_SPEC

Key for the model spec.

See Also:

Constant Field Values

CLASS_COLUMN

public static final DataColumnSpec CLASS_COLUMN

Name of the basisfunction class column.

Constructor Detail

BasisFunctionFactory

protected BasisFunctionFactory(DataTableSpec spec,
                               String[] targetColumns,
                               DataType type,
                               int distance)

Creates new basisfunction factory with the given spec to extract min/max value for all numeric columns.

Parameters:

spec - the training's data spec

targetColumns - the class info column in the data

type - the type for the model columns

distance - the choice of distance function

Method Detail

findDataColumns

public static final String[] findDataColumns(DataTableSpec spec,
                                             List<String> targetCols)

Find all numeric columns which are not target columns.

Parameters:

spec - the input spec

targetCols - column(s) set as target

Returns:

array of data column names

getMinimums

public final MutableDouble[] getMinimums()

Returns:

the lower bounds.

getMaximums

public final MutableDouble[] getMaximums()

Returns:

the upper bounds.

createModelSpec

public static final DataTableSpec createModelSpec(DataTableSpec inSpec,
                                                  String[] dataColumns,
                                                  String[] targetColumns,
                                                  DataType type)

Creates a model spec based on the data input spec by extracting all DoubleCell columns and the specified target column.

Parameters:

inSpec - the input data spec

dataColumns - the data columns used for training

targetColumns - the target classification columns

type - the type for the model columns

Returns:

a new table spec with a number of DoubleCells and the target column last

getDistance

public final int getDistance()

Returns the choice of distance function.

Returns:

distance function

getModelSpec

public DataTableSpec getModelSpec()

Returns:

the model's spec with class info column

commit

public abstract BasisFunctionLearnerRow commit(RowKey key,
                                               DataCell classInfo,
                                               DataRow row)

Returns a new row initialised by a DataRow as its initial center vector and a class label.

Parameters:

key - this row's key

classInfo - data cell contains class info

row - the initial center vector

Returns:

a new row of a certain type

save

public void save(ModelContent pp)

Saves to model content.

Parameters:

pp - the model content this is saved to.