Package

weka.filters.unsupervised.attribute

Synopsis

Converts the given set of predictor variables into a kernel matrix. The class value remains unchanged, as long as the preprocessing filter doesn't change it.
By default, the data is preprocessed with the Center filter, but the user can choose any filter (NB: one must be careful that the filter does not alter the class attribute unintentionally). With weka.filters.AllFilter the preprocessing gets disabled.

For more information regarding preprocessing the data, see:

K.P. Bennett, M.J. Embrechts: An Optimization Perspective on Kernel Partial Least Squares Regression. In: Advances in Learning Theory: Methods, Models and Applications, 227-249, 2003.

Options

The table below describes the options available for KernelFilter.

Option	Description
checksTurnedOff	Turns time-consuming checks off - use with caution.
debug	Turns on output of debugging information.
initFile	The dataset to initialize the filter with.
initFileClassIndex	The class index of the dataset to initialize the filter with (first and last are valid).
kernel	The kernel to use.
kernelFactorExpression	The factor for the kernel, with A = # of attributes and N = # of instances.
preprocessing	Sets the filter to use for preprocessing (use the AllFilter for no preprocessing).

Capabilities

The table below describes the capabilites of KernelFilter.

Capability	Supported
Class	Relational class, Unary class, Binary class, String class, Missing class values, Date class, Nominal class, Numeric class, Empty nominal class
Attributes	Numeric attributes
Min # of instances	0

Browser not supported