NAME

Algorithm::KernelKMeans::Util

DESCRIPTION

This module provides some constants and functions suitable to use with Algorithm::KernelKMeans.

CONSTANTS

Constants listed below represent kernels/initializers, which some methods require. It's recommended to use these constants instead of code reference, because clusterer implementations might have its own kernel/initializer code. So code references should be used iff there's no equivalent constants.

All constants are import()able.

$KERNEL_POLYNOMINAL

Polynominal kernel. Takes 2 parameters ($l, $p) then will be formed like K(x1, x2) = ($l + x1 . x2)^$p, where "x1 . x2" represents inner product.

$KERNEL_GAUSSIAN

Gaussian kernel. Takes 1 parameter ($sigma).

K(x1, x2) = exp(-||x1 - x2||^2 / (2 * $sigma)^2)

$KERNEL_SIGMOID

Sigmoid kernel. Takes 2 parameters ($s, $theta).

K(x1, x2) = tanh($s * (x1 . x2) + $theta)

$INITIALIZE_SIMPLE

$INITIALIZE_SHUFFLE

$INITIALIZE_KKZ

FUNCTIONS

This module exports nothing by default. You can import functions below:

centroid($cluster)

Takes array ref of vectors and returns centroid vector of the cluster.

inner_product($v, $u)

Calculates inner product of $v and $u.

euclidean_distance($v, $u)

Computes euclidean distance between $v and $u.

AUTHOR

Koichi SATOH <r.sekia@gmail.com>