% 
% Written by:
% -- 
% John L. Weatherwax                2005-08-14
% 
% email: wax@alum.mit.edu
% 
% Please send comments and especially bug reports to the
% above email address.
% 
%-----

close all; drawnow; 
clc; 
clear; addpath( '../../Code/eda_data' ); addpath( '../../Code/eda_toolbox' ); addpath( '../Chapter1' );

% Example 8.4
load L1bpm
% Get just those topics on 5 through 11
ind = find(classlab == 5);
ind = [ind; find(classlab == 6)];
ind = [ind; find(classlab == 8)];
ind = [ind; find(classlab == 9)];
ind = [ind; find(classlab == 11)];
% Change the class labels for class 11
% for better plotting.
clabs = classlab(ind);
clabs(find(clabs==11)) = 1;

% First do dimensionality reduction - ISOMAP
[Y, R, E] = isomap(L1bpm,'k',10);
% Choose 3 dimensions, based on residual variance.
XX = Y.coords{3}';
% Only need those observations in classes of interest.
X = XX(ind,:);
% Now do model-based clustering with a maximum of 10 clusters.
[bics,bestm,allm,Z,clabsmbc] = mbclust(X,10);

% Get the hierarchical clustering.
Y = pdist(X);
Z = linkage(Y,'complete');
dendrogram(Z);
cids = cluster(Z,'maxclust',16);            % <- we now ask for 16 groups 

% An exension of Example 8.7
% Applies Ling's method to interpoint distance matrix.

if( 0 ) 
  % Continuing with same data used in  Example 8.4. but Use just Topics 6 and 9
  ind = find(classlab == 6);
  ind = [ind; find(classlab == 9)];
  clabs = classlab(ind);
else
  ind = 1:size(XX,1);
end
n = length(ind);
data = XX(ind,:);
% Randomly reorder the data and view the interpoint distance matrix as an image.
data = data(randperm(n),:);
Y = pdist(data);
Ys = squareform(Y);
imagesc(Ys);
colormap(gray(256))
title('Interpoint Distance Matrix - Random Order')
axis off
saveas( gcf, '../../WriteUp/Graphics/Chapter8/prob_8_6_ro', 'epsc' ); 

% Now apply Ling's method. First need to get a partition or clustering.
Z = linkage(Y,'complete');
% Now get the ordering based on the cluster scheme.
T = cluster(Z,'maxclust',16);
% Sort these, so points in the same cluster are adjacent.
[Ts,inds] = sort(T);
% Sort the distance matrix using this and replot.
figure
imagesc(Ys(inds,inds))
title('Interpoint Distance Matrix - Cluster Order')
colormap(gray(256))
axis off
saveas( gcf, '../../WriteUp/Graphics/Chapter8/prob_8_6_co', 'epsc' );