org.knime.base.node.mine.cluster.fuzzycmeans

Class FCMAlgorithmMemory

java.lang.Object

org.knime.base.node.mine.cluster.fuzzycmeans.FCMAlgorithm

org.knime.base.node.mine.cluster.fuzzycmeans.FCMAlgorithmMemory

public class FCMAlgorithmMemory
extends FCMAlgorithm

The Fuzzy c-means algorithm.

Author:

Nicolas Cebron, University of Konstanz

Constructor Summary

Constructors

Constructor and Description
FCMAlgorithmMemory(int nrClusters, double fuzzifier) Constructor for a Fuzzy c-means algorithm (with no noise detection).
FCMAlgorithmMemory(int nrClusters, double fuzzifier, boolean calculateDelta, double deltalambda) Constructor for a Fuzzy c-means algorithm with noise detection.

Method Summary

Modifier and Type	Method and Description
double	doOneIteration(ExecutionContext exec) Does one iteration in the Fuzzy c-means algorithm.
double[][]	getConvertedData() Please make sure to call init() first in order to guarantee that the DataTable is converted.
double[][]	getConvertedData(DataTable table)
double[][]	getConvertedData(DataTable table, Random random)
RowKey[]	getRowKeys() Please make sure to call init() first in order to guarantee that the DataTable is converted.
void	init(int nrRows, int dimension, DataTable table) Inits the cluster centers and the weight matrix.
void	init(int nrRows, int dimension, DataTable table, Random random) Inits the cluster centers and the weight matrix.
void	init(RowKey[] keys, double[][] data) Inits the cluster centers and the weight matrix.
void	init(RowKey[] keys, double[][] data, Random random) Inits the cluster centers and the weight matrix.

Methods inherited from class org.knime.base.node.mine.cluster.fuzzycmeans.FCMAlgorithm

addTotalChange, getClusterCentres, getClusters, getDelta, getDimension, getFuzzifier, getLambda, getNrClusters, getNrRows, getTotalChange, getweightMatrix, getWeightMatrix, getWinner, init, init, isCalculateDelta, isNoise, noiseClustering, setClusterValue, setDelta, setTotalChange, setWeightMatrixValue

Methods inherited from class java.lang.Object

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

Constructor Detail

FCMAlgorithmMemory

public FCMAlgorithmMemory(int nrClusters,
                          double fuzzifier)

Constructor for a Fuzzy c-means algorithm (with no noise detection).

Parameters:

nrClusters - the number of cluster prototypes to use

fuzzifier - allows the clusters to overlap

FCMAlgorithmMemory

public FCMAlgorithmMemory(int nrClusters,
                          double fuzzifier,
                          boolean calculateDelta,
                          double deltalambda)

Constructor for a Fuzzy c-means algorithm with noise detection. It can be indicated, whether the delta value of the noise cluster should be updated automatically or if it should be calculated automatically. The last parameter specifies either the delta value or the lambda value, depending on the boolean flag in the parameter before.

Parameters:

nrClusters - the number of clusters to use

fuzzifier - the fuzzifier, controls how much the clusters can overlap

calculateDelta - indicate whether delta should be calculated automatically

deltalambda - the delta value, if the previous parameter is false, the lambda value otherwise

Method Detail

init

public void init(int nrRows,
                 int dimension,
                 DataTable table)

Inits the cluster centers and the weight matrix. Must be called before the iterations are carried out.

Overrides:

init in class FCMAlgorithm

Parameters:

nrRows - number of rows in the DataTable

dimension - the dimension of the table

table - the table to use.

init

public void init(int nrRows,
                 int dimension,
                 DataTable table,
                 Random random)

Inits the cluster centers and the weight matrix. Must be called before the iterations are carried out.

Overrides:

init in class FCMAlgorithm

Parameters:

nrRows - number of rows in the DataTable

dimension - the dimension of the table

table - the table to use

random - a random number generator for initializing the cluster centers

Since:

2.8

init

public void init(RowKey[] keys,
                 double[][] data)

Inits the cluster centers and the weight matrix. Must be called before the iterations are carried out.

Parameters:

keys - the RowKeys for each data row.

data - the DaaTable as 2 dimensional double array.

init

public void init(RowKey[] keys,
                 double[][] data,
                 Random random)

Inits the cluster centers and the weight matrix. Must be called before the iterations are carried out.

Parameters:

keys - the RowKeys for each data row

data - the DaaTable as 2 dimensional double array

random - a random number generator for initializing the cluster centers

Since:

2.8

getConvertedData

public double[][] getConvertedData()

Please make sure to call init() first in order to guarantee that the DataTable is converted.

Returns:

the input DataTable converted as a double array. If it has not been produced yet, null is returned.

getConvertedData

public double[][] getConvertedData(DataTable table)

Parameters:

table - DataTable to convert.

Returns:

two-dimensional double array.

getConvertedData

public double[][] getConvertedData(DataTable table,
                                   Random random)

Parameters:

table - DataTable to convert

random - a random number generator for initializing the cluster centers

Returns:

two-dimensional double array

Since:

2.8

getRowKeys

public RowKey[] getRowKeys()

Please make sure to call init() first in order to guarantee that the DataTable is converted.

Returns:

the RowKeys assigned to each position of the produced double array

See Also:

getConvertedData()

doOneIteration

public double doOneIteration(ExecutionContext exec)
                      throws CanceledExecutionException

Does one iteration in the Fuzzy c-means algorithm. First, the weight matrix is updated and then the cluster prototypes are recalculated.

Overrides:

doOneIteration in class FCMAlgorithm

Parameters:

exec - execution context to cancel the execution

Returns:

the total change in the cluster prototypes. Allows to decide whether the algorithm can be stopped.

Throws:

CanceledExecutionException - if the operation is canceled