KNN

The KNN module implements a K Nearest Neighbors System. It has inputs for simultaneous learning and choosing of neighbors. The underlying data structure used is a KD-Tree.

Example XML definition

A simple example

  <module
      class = "KNN"
      name = "KNN"
  />

Parameters

Name	Description	Type	Default
class	The class name for the module; must be "KNN".	string
name	The name of this instance of the module.	string
k	How many neighbors to look for.	int	5
distance_type	Type of distance calculation. Euclidian is the usual sqrt(sum of squares), manhattan is just the addition of differences in all the dimensions.	choices: euclidian manhattan	euclidian
auto_rebuild	Automatically check if some sub tree can be balanced after an insert.	bool	true
minimum_tree_size_for_rebuild	Will not auto rebuild smaller sub trees than this.	int	100
unbalanced_tree_size_ratio_limit	Defines what is considered unbalanced. The ratio means ratio between the sizes of the left and right sub trees of a tree.	float	2.4
verbose	Print some info.	bool	false
check_for_clones	Dont insert an example if it is already in the tree.	bool	false

Module Connections

Inputs

Name	Description
INPUT	The coordinates of a target to look for neighbors for.
T_INPUT	Training input. The coordinates of an example to learn/store.
T_OUTPUT	Training output. The class values of the example to learn/store
LEARN	Learns from T_INPUT/T_OUTPUT if this is > 0.

Outputs

Name	Description
INPUT_TABLE	A matrix of the coordinates of the neighbors that were found around the last INPUT.
OUTPUT_TABLE	A matrix of the class values of the neighbors that were found around the last INPUT.
DISTANCE_TABLE	An array of the distances of the neighbors that were found around the last INPUT. NOTE: When it was not possible to find K neighbors (the tree was smaller than size K nodes) the distances for the neighbors that were not found is set to -1.0

Author

Alexander Kolodziej
LUCS

Files

KNN.h
KNN.cc
KD_Tree.ccc
KD_Tree.h
KNN.ikc

KNN