distMatrixShow

PURPOSE ^

Useful visualization of a distance matrix of clustered points.

SYNOPSIS ^

function [D, Dsm] = distMatrixShow( D, IDX, show )

DESCRIPTION ^

 Useful visualization of a distance matrix of clustered points.

 D is sorted into k blocks, where the ith block contains all the points in
 cluster i. When D is displayed the blocks are shown explicitly.  Hence
 for a good clustering (under a spherical gaussian assumption) the
 'diagonal' blocks ought to be mostly dark, and all other block ought to be
 relatively white.  One can thus quickly visualize the quality of the
 clustering, or even how clusterable the points are.  Outliers (according
 to IDX) are removed from D.

 USAGE
  [D, Dsm] = distMatrixShow( D, IDX, [show] )

 INPUTS
  D       - nxn distance matrix
  IDX     - cluster membership [see kmeans2.m]
  show    - [1] will display results in figure(show)

 OUTPUTS
  D       - sorted nxn distance matrix
  Dsm     - sorted and smoothed nxn distance matrix

 EXAMPLE
  % not the best example since points are already ordered
  [X,IDX] = demoGenData(100,0,5,2,10,2,0);
  distMatrixShow( pdist2(X,X), IDX );

 See also VISUALIZEDATA, KMEANS2

 Piotr's Computer Vision Matlab Toolbox      Version 2.0
 Copyright 2014 Piotr Dollar.  [pdollar-at-gmail.com]
 Licensed under the Simplified BSD License [see external/bsd.txt]

Generated by m2html © 2003