som_demo3.m

很全的som工具箱 四个demo可为初学者提够帮助

pause % Strike any key to continue...

%    Besides SOM_SHOW and SOM_CPLANE, there are three other
%    functions specifically designed for showing the values of the 
%    component planes: SOM_PIEPLANE, SOM_BARPLANE, SOM_PLOTPLANE. 

%    SOM_PIEPLANE shows a single pie chart for each map unit.  Each
%    pie shows the relative proportion of each component of the sum of
%    all components in that map unit. The component values must be
%    positive. 

%    SOM_BARPLANE shows a barchart in each map unit. The scaling of 
%    bars can be made unit-wise or variable-wise. By default it is
%    determined variable-wise.

%    SOM_PLOTPLANE shows a linegraph in each map unit. 

M = som_normalize(sM.codebook,'range'); 

subplot(1,3,1)
som_pieplane(sM, M);
title('som\_pieplane')

subplot(1,3,2)
som_barplane(sM, M, '', 'unitwise');
title('som\_barplane')

subplot(1,3,3)
som_plotplane(sM, M, 'b');
title('som\_plotplane')

pause % Strike any key to visualize cluster properties...



clf
clc
%    2. VISUALIZATION OF COMPONENTS: CLUSTERS
%    ========================================

%    An interesting question is of course how do the values of the
%    variables relate to the clusters: what are the values of the
%    components in the clusters, and which components are the ones
%    which *make* the clusters.

som_show(sM)

%    From the U-matrix and component planes, one can easily see
%    what the typical values are in each cluster. 

pause % Strike any key to continue...

%    The significance of the components with respect to the clustering
%    is harder to visualize. One indication of importance is that on
%    the borders of the clusters, values of important variables change
%    very rabidly.

%    Here, the distance matrix is calculated with respect to each
%    variable. 

u1 = som_umat(sM,'mask',[1 0 0]'); u1=u1(1:2:size(u1,1),1:2:size(u1,2));
u2 = som_umat(sM,'mask',[0 1 0]'); u2=u2(1:2:size(u2,1),1:2:size(u2,2));
u3 = som_umat(sM,'mask',[0 0 1]'); u3=u3(1:2:size(u3,1),1:2:size(u3,2));

%    Here, the distance matrices are shown, as well as a piechart
%    indicating the relative importance of each variable in each
%    map unit. The size of piecharts has been scaled by the
%    distance matrix calculated from all components.

subplot(2,2,1)
som_cplane(sM,u1(:));
title(sM.comp_names{1})

subplot(2,2,2)
som_cplane(sM,u2(:));
title(sM.comp_names{2})

subplot(2,2,3)
som_cplane(sM,u3(:));
title(sM.comp_names{3})

subplot(2,2,4)
som_pieplane(sM, [u1(:), u2(:), u3(:)], hsv(3), Um(:)/max(Um(:)));
title('Relative importance')

%    From the last subplot, one can see that in the area where the
%    bigger cluster border is, the 'X-coord' component (red color)
%    has biggest effect, and thus is the main factor in separating
%    that cluster from the rest.

pause % Strike any key to learn about correlation hunting...


clf
clc
%    2. VISUALIZATION OF COMPONENTS: CORRELATION HUNTING
%    ===================================================

%    Finally, the component planes are often used for correlation
%    hunting. When the number of variables is high, the component
%    plane visualization offers a convenient way to visualize all
%    components at once and hunt for correlations (as opposed to
%    N*(N-1)/2 scatterplots).

%    Hunting correlations this way is not very accurate. However, it
%    is easy to select interesting combinations for further
%    investigation.

%    Here, the first and third components are shown with scatter
%    plot. As with projections, a color coding is used to link the
%    visualization to the map plane. In the color coding, size shows
%    the distance matrix information.

C = som_colorcode(sM);
subplot(1,2,1)
som_cplane(sM,C,1-Um(:)/max(Um(:)));
title('Color coding + distance matrix')

subplot(1,2,2)
som_grid(sM,'Coord',sM.codebook(:,[1 3]),'MarkerColor',C);
title('Scatter plot'); xlabel(sM.comp_names{1}); ylabel(sM.comp_names{3})
axis equal

pause % Strike any key to visualize data responses...


clf
clc
%    3. DATA ON MAP
%    ==============

%    The SOM is a map of the data manifold. An interesting question
%    then is where on the map a specific data sample is located, and
%    how accurate is that localization? One is interested in the
%    response of the map to the data sample. 

%    The simplest answer is to find the BMU of the data sample.
%    However, this gives no indication of the accuracy of the
%    match. Is the data sample close to the BMU, or is it actually
%    equally close to the neighboring map units (or even approximately
%    as close to all map units)? Sometimes accuracy doesn't really
%    matter, but if it does, it should be visualized somehow.

%    Here are different kinds of response visualizations for two
%    vectors: [0 0 0] and [99 99 99]. 
%     - BMUs indicated with labels       
%     - BMUs indicated with markers, relative quantization errors
%       (in this case, proportion between distances to BMU and 
%       Worst-MU) with vertical lines
%     - quantization error between the samples and all map units 
%     - fuzzy response (a non-linear function of quantization
%       error) of all map units

echo off
[bm,qe] = som_bmus(sM,[0 0 0; 99 99 99],'all'); % distance to all map units
[dummy,ind] = sort(bm(1,:)); d0 = qe(1,ind)'; 
[dummy,ind] = sort(bm(2,:)); d9 = qe(2,ind)'; 
bmu0 = bm(1,1); bmu9 = bm(2,1); % bmus

h0 = zeros(prod(sM.topol.msize),1); h0(bmu0) = 1; % crisp hits
h9 = zeros(prod(sM.topol.msize),1); h9(bmu9) = 1; 

lab = cell(prod(sM.topol.msize),1); 
lab{bmu0} = '[0,0,0]'; lab{bmu9} = '[99,99,99]';

hf0 = som_hits(sM,[0 0 0],'fuzzy'); % fuzzy response
hf9 = som_hits(sM,[99 99 99],'fuzzy'); 

som_show(sM,'umat',{'all','BMU'},...
	 'color',{d0,'Qerror 0'},'color',{hf0,'Fuzzy response 0'},...
	 'empty','BMU+qerror',...
	 'color',{d9,'Qerror 99'},'color',{hf9,'Fuzzy response 99'});	 
som_show_add('label',lab,'Subplot',1,'Textcolor','r');
som_show_add('hit',[h0, h9],'Subplot',4,'MarkerColor','r');
hold on
Co = som_vis_coords(sM.topol.lattice,sM.topol.msize);
plot3(Co(bmu0,[1 1]),Co(bmu0,[2 2]),[0 10*qe(1,1)/qe(1,end)],'r-')
plot3(Co(bmu9,[1 1]),Co(bmu9,[2 2]),[0 10*qe(2,1)/qe(2,end)],'r-')
view(3), axis equal
echo on

%    Here are the distances to BMU, 2-BMU and WMU:

qe(1,[1,2,end]) % [0 0 0]
qe(2,[1,2,end]) % [99 99 99]

%    One can see that for [0 0 0] the accuracy is pretty good as the
%    quantization error of the BMU is much lower than that of the
%    WMU. On the other hand [99 99 99] is very far from the map:
%    distance to BMU is almost equal to distance to WMU.

pause % Strike any key to visualize responses of multiple samples...



clc
clf
%    3. DATA ON MAP: HIT HISTOGRAMS
%    ==============================

%    One can also investigate whole data sets using the map. When the
%    BMUs of multiple data samples are aggregated, a hit histogram
%    results. Instead of BMUs, one can also aggregate for example
%    fuzzy responses.

%    The hit histograms (or aggregated responses) can then be compared
%    with each other. 

%    Here are hit histograms of three data sets: one with 50 first
%    vectors of the data set, one with 150 samples from the data
%    set, and one with 50 randomly selected samples. In the last
%    subplot, the fuzzy response of the first data set.

dlen = size(sD.data,1);
Dsample1 = sD.data(1:50,:); h1 = som_hits(sM,Dsample1); 
Dsample2 = sD.data(1:150,:); h2 = som_hits(sM,Dsample2); 
Dsample3 = sD.data(ceil(rand(50,1)*dlen),:); h3 = som_hits(sM,Dsample3); 
hf = som_hits(sM,Dsample1,'fuzzy');

som_show(sM,'umat','all','umat','all','umat','all','color',{hf,'Fuzzy'})
som_show_add('hit',h1,'Subplot',1,'Markercolor','r')
som_show_add('hit',h2,'Subplot',2,'Markercolor','r')
som_show_add('hit',h3,'Subplot',3,'Markercolor','r')

pause % Strike any key to visualize trajectories...



clc
clf
%    3. DATA ON MAP: TRAJECTORIES
%    ============================

%    A special data mapping technique is trajectory. If the samples
%    are ordered, forming a time-series for example, their response on
%    the map can be tracked. The function SOM_SHOW_ADD can be used to
%    show the trajectories in two different modes: 'traj' and 'comet'.

%    Here, a series of data points is formed which go from [8,0,0]
%    to [2,2,2]. The trajectory is plotted using the two modes.

Dtraj = [linspace(9,2,20); linspace(0,2,20); linspace(0,2,20)]';
T = som_bmus(sM,Dtraj);

som_show(sM,'comp',[1 1]);
som_show_add('traj',T,'Markercolor','r','TrajColor','r','subplot',1);
som_show_add('comet',T,'MarkerColor','r','subplot',2);

%    There's also a function SOM_TRAJECTORY which lauches a GUI
%    specifically designed for displaying trajectories (in 'comet'
%    mode).

pause % Strike any key to learn about color handling...




clc
clf
%    COLOR HANDLING
%    ==============

%    Matlab offers flexibility in the colormaps. Using the COLORMAP
%    function, the colormap may be changed. There are several useful
%    colormaps readily available, for example 'hot' and 'jet'. The
%    default number of colors in the colormaps is 64. However, it is
%    often advantageous to use less colors in the colormap. This way
%    the components planes visualization become easier to interpret.

%    Here the three component planes are visualized using the 'hot'
%    colormap and only three colors.

som_show(sM,'comp',[1 2 3])
colormap(hot(3));
som_recolorbar 

pause % Press any key to change the colorbar labels...

%     The function SOM_RECOLORBAR can be used to reconfigure
%     the labels beside the colorbar. 

%     Here the colorbar of the first subplot is labeled using labels
%     'small', 'medium' and 'big' at values 0, 1 and 2.  For the
%     colorbar of the second subplot, values are calculated for the
%     borders between colors.

som_recolorbar(1,{[0 4 9]},'',{{'small','medium','big'}});
som_recolorbar(2,'border','');

pause % Press any key to learn about SOM_NORMCOLOR...

%     Some SOM Toolbox functions do not use indexed colors if the
%     underlying Matlab function (e.g. PLOT) do not use indexed
%     colors. SOM_NORMCOLOR is a convenient function to simulate
%     indexed colors: it calculates fixed RGB colors that
%     are similar to indexed colors with the specified colormap. 

%     Here, two SOM_GRID visualizations are created. One uses the
%     'surf' mode to show the component colors in indexed color
%     mode, and the other uses SOM_NORMALIZE to do the same. 

clf
colormap(jet(64))
subplot(1,2,1)
som_grid(sM,'Surf',sM.codebook(:,3));
title('Surf mode')

subplot(1,2,2)
som_grid(sM,'Markercolor',som_normcolor(sM.codebook(:,3)));
title('som\_normcolor')

pause % Press any key to visualize different map shapes...



clc
clf
%    DIFFERENT MAP SHAPES
%    ====================

%    There's no direct way to visualize cylinder or toroid maps. When
%    visualized, they are treated exactly as if they were sheet
%    shaped. However, if function SOM_UNIT_COORDS is used to provide
%    unit coordinates, then SOM_GRID can be used to visualize these
%    alternative map shapes.

%    Here the grids of the three possible map shapes (sheet, cylinder
%    and toroid) are visualized. The last subplot shows a component 
%    plane visualization of the toroid map.

Cor = som_unit_coords(sM.topol.msize,'hexa','sheet');
Coc = som_unit_coords(sM.topol.msize,'hexa','cyl');
Cot = som_unit_coords(sM.topol.msize,'hexa','toroid');

subplot(2,2,1)
som_grid(sM,'Coord',Cor,'Markersize',3,'Linecolor','k');
title('sheet'), view(0,-90), axis tight, axis equal

subplot(2,2,2)
som_grid(sM,'Coord',Coc,'Markersize',3,'Linecolor','k');
title('cylinder'), view(5,1), axis tight, axis equal

subplot(2,2,3)
som_grid(sM,'Coord',Cot,'Markersize',3,'Linecolor','k');
title('toroid'), view(-100,0), axis tight, axis equal

subplot(2,2,4)
som_grid(sM,'Coord',Cot,'Surf',sM.codebook(:,3));
colormap(jet), colorbar
title('toroid'), view(-100,0), axis tight, axis equal

echo off