package jsat.datatransform;
import jsat.DataSet;
import jsat.classifiers.DataPoint;
import jsat.linear.Vec;
/**
* This transform performs standardization of the data, which makes each column
* have a mean of zero and a variance of one. This assume the data comes from a
* normal distribution and scales it to the unit normal distribution.
*
* This transform is equivalent to applying {@link ZeroMeanTransform} followed
* by {@link UnitVarianceTransform}.
*
* @author Edward Raff
*/
public class StandardizeTransform implements InPlaceTransform
{
private static final long serialVersionUID = -2349721113741805955L;
private Vec means;
private Vec stdDevs;
/**
* Creates a new object for Standardizing datasets
*/
public StandardizeTransform()
{
}
/**
* Creates a new object for standaidizing datasets fit to the given dataset
* @param dataset the dataset to learn how to standardize from
*/
public StandardizeTransform(DataSet dataset)
{
fit(dataset);
}
@Override
public void fit(DataSet dataset)
{
Vec[] vecs = dataset.getColumnMeanVariance();
means = vecs[0];
stdDevs = vecs[1];
for(int i = 0; i < stdDevs.length(); i++)//convert variant to stndDev
stdDevs.set(i, Math.sqrt(stdDevs.get(i)+1e-10));
}
/**
* Copy constructor
* @param toCopy the object to copy
*/
public StandardizeTransform(StandardizeTransform toCopy)
{
this.means = toCopy.means.clone();
this.stdDevs = toCopy.stdDevs.clone();
}
@Override
public DataPoint transform(DataPoint dp)
{
DataPoint newDP = dp.clone();
mutableTransform(newDP);
return newDP;
}
@Override
public void mutableTransform(DataPoint dp)
{
Vec toAlter = dp.getNumericalValues();
toAlter.mutableSubtract(means);
toAlter.mutablePairwiseDivide(stdDevs);
}
@Override
public boolean mutatesNominal()
{
return false;
}
@Override
public StandardizeTransform clone()
{
return new StandardizeTransform(this);
}
}
Ads help maintain this website.