databionics.esom.train
Class BatchSOM

java.lang.Object
  databionics.esom.train.SOM
      databionics.esom.train.BatchSOM

public class BatchSOM
extends SOM

Implementation of the batch SOM training. The update of the bestmatches and its neighbors is done after each epoch.

Field Summary

protected cern.colt.map.OpenIntObjectHashMap	hashMap Hash map to connect bm index and bm hitcount
protected static org.apache.log4j.Logger	log interface to log4j system

Fields inherited from class databionics.esom.train.SOM

bestmatches, bmSearch, bmStat, bmStatNeeded, center, classes, count, data, descriptives, distanceFunction, distances, epoch, epochs, grid, initMethod, keys, lrn, neighborhood, neuron, nf, offline, oldBestmatches, online, pattern, permutation, permutations, permute, qerror, radius, radiusCooling, random, rate, rateCooling, saveEpoch, saveEpoches, savePrefix, saveUMatrix, view

Constructor Summary

BatchSOM()
Standard constructor

Method Summary

void	afterEpoch() Now, the neurons are updated.
void	afterUpdate(int index, int row) Store the data vector projected on this bestmatch in a BestMatchHit object.
void	update(cern.colt.matrix.DoubleMatrix1D vector, int bm, int pos) Empty implementation.
void	updateNeighborhood(cern.colt.matrix.DoubleMatrix1D vector, int bm) Update a neuron and its neighborhood.

Methods inherited from class databionics.esom.train.SOM

beforeEpoch, beforeSearch, beforeUpdate, calcQerror, centerMap, cool, getBestMatches, getBmSearch, getData, getDescriptives, getDistance, getDistance, getDistanceFunction, getDistances, getEpochs, getGrid, getKeys, getNeighborhood, getNumberFormat, getOldBestmatches, getPattern, getPermutation, getPermute, getRadius, getRadiusCooling, getRateCooling, init, isBmStatNeeded, isCenter, loadCls, loadData, saveBestMatches, setBestMatches, setBestMatches, setBmSearch, setBmStatNeeded, setCenter, setData, setDistanceFunction, setDistances, setEpochs, setGrid, setInit, setKeys, setNeighborhood, setNumberFormat, setOldBestmatches, setOnline, setPermute, setRadiusCooling, setRateCooling, setSaveEpoch, setSaveEpochBoolean, setSavePrefix, setSaveUMatrix, stop, train, updateNeuron

Methods inherited from class java.lang.Object

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

Field Detail

log

protected static final org.apache.log4j.Logger log

interface to log4j system

hashMap

protected cern.colt.map.OpenIntObjectHashMap hashMap

Hash map to connect bm index and bm hitcount

Constructor Detail

BatchSOM

public BatchSOM()

Standard constructor

Method Detail

updateNeighborhood

public void updateNeighborhood(cern.colt.matrix.DoubleMatrix1D vector,
                               int bm)

Update a neuron and its neighborhood.

Overrides:

updateNeighborhood in class SOM

Parameters:

vector - vector of data

bm - index of bestmatching neuron

update

public void update(cern.colt.matrix.DoubleMatrix1D vector,
                   int bm,
                   int pos)

Empty implementation. Nothing to do in batch mode.

Specified by:

update in class SOM

Parameters:

vector - vector of data

bm - index of bestmatching neuron

afterUpdate

public void afterUpdate(int index,
                        int row)

Store the data vector projected on this bestmatch in a BestMatchHit object.

Overrides:

afterUpdate in class SOM

Parameters:

index - Index of neuron to be updated

row - Index of pattern in data

afterEpoch

public void afterEpoch()

Now, the neurons are updated.

Overrides:

afterEpoch in class SOM