Mercurial > hg > camir-aes2014

annotate toolboxes/FullBNT-1.0.7/bnt/examples/static/HME/hme_class_plot.m @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
rev   line source
wolffd@0 1 function fh=hme_class_plot(net, nodes_info, train_data, test_data)
wolffd@0 2 %
wolffd@0 3 % Use this function ONLY when the input dimension is 2
wolffd@0 4 % and the problem is a classification one.
wolffd@0 5 % We assume that each row of 'train_data' & 'test_data' is an example.
wolffd@0 6 %
wolffd@0 7 %------Line Spec------------------------------------------------------------------------
wolffd@0 8 %
wolffd@0 9 % LineWidth - specifies the width (in points) of the line
wolffd@0 10 % MarkerEdgeColor - specifies the color of the marker or the edge color
wolffd@0 11 % forfilled markers (circle, square, diamond, pentagram, hexagram, and the
wolffd@0 12 % four triangles).
wolffd@0 13 % MarkerFaceColor - specifies the color of the face of filled markers.
wolffd@0 14 % MarkerSize - specifies the size of the marker in points.
wolffd@0 15 %
wolffd@0 16 % Example
wolffd@0 17 % -------
wolffd@0 18 % plot(t,sin(2*t),'-mo',...
wolffd@0 19 % 'LineWidth',2,...
wolffd@0 20 % 'MarkerEdgeColor','k',... % 'k'=black
wolffd@0 21 % 'MarkerFaceColor',[.49 1 .63],... % RGB color
wolffd@0 22 % 'MarkerSize',12)
wolffd@0 23 %----------------------------------------------------------------------------------------
wolffd@0 24
wolffd@0 25 class_num=nodes_info(2,end);
wolffd@0 26 mn_x = round(min(train_data(:,1))); mx_x = round(max(train_data(:,1)));
wolffd@0 27 mn_y = round(min(train_data(:,2))); mx_y = round(max(train_data(:,2)));
wolffd@0 28 if nargin==4,
wolffd@0 29 mn_x = round(min([train_data(:,1); test_data(:,1)]));
wolffd@0 30 mx_x = round(max([train_data(:,1); test_data(:,1)]));
wolffd@0 31 mn_y = round(min([train_data(:,2); test_data(:,2)]));
wolffd@0 32 mx_y = round(max([train_data(:,1); test_data(:,2)]));
wolffd@0 33 end
wolffd@0 34 x = mn_x(1)-1:0.2:mx_x(1)+1;
wolffd@0 35 y = mn_y(1)-1:0.2:mx_y(1)+1;
wolffd@0 36 [X, Y] = meshgrid(x,y);
wolffd@0 37 X = X(:);
wolffd@0 38 Y = Y(:);
wolffd@0 39 num_g=size(X,1);
wolffd@0 40 griglia = [X Y];
wolffd@0 41 rand('state',1);
wolffd@0 42 if class_num<=6,
wolffd@0 43 colors=['r'; 'g'; 'b'; 'c'; 'm'; 'y'];
wolffd@0 44 else
wolffd@0 45 colors=rand(class_num, 3); % each row is an RGB color
wolffd@0 46 end
wolffd@0 47 fh = figure('Name','Data & decision boundaries', 'MenuBar', 'none', 'NumberTitle', 'off');
wolffd@0 48 ms=5; % Marker Size
wolffd@0 49 if nargin==4,
wolffd@0 50 % ms=4; % Marker Size
wolffd@0 51 subplot(1,2,1);
wolffd@0 52 end
wolffd@0 53 % Plot of train_set -------------------------------------------------------------------------
wolffd@0 54 axis([mn_x-1 mx_x+1 mn_y-1 mx_y+1]);
wolffd@0 55 set(gca, 'Box', 'on');
wolffd@0 56 c_max_train = max(train_data(:,3));
wolffd@0 57 hold on
wolffd@0 58 for m=1:c_max_train,
wolffd@0 59 app_x=train_data(:,1);
wolffd@0 60 app_y=train_data(:,2);
wolffd@0 61 thisX=app_x(train_data(:,3)==m);
wolffd@0 62 thisY=app_y(train_data(:,3)==m);
wolffd@0 63 if class_num<=6,
wolffd@0 64 str_col=[];
wolffd@0 65 str_col=['o', colors(m,:)];
wolffd@0 66 plot(thisX, thisY, str_col, 'MarkerSize', ms);
wolffd@0 67 else
wolffd@0 68 plot(thisX, thisY, 'o',...
wolffd@0 69 'LineWidth', 1,...
wolffd@0 70 'MarkerEdgeColor', colors(m,:), 'MarkerSize', ms)
wolffd@0 71 end
wolffd@0 72 end
wolffd@0 73 %---hmefwd_generale(net,data,ndata)-----------------------------------------------------------
wolffd@0 74 Z=fhme(net, nodes_info, griglia, num_g); % forward propagation trougth the HME
wolffd@0 75 %---------------------------------------------------------------------------------------------
wolffd@0 76 [foo , class] = max(Z'); % 0/1 loss function => we assume that the true class is the one with the
wolffd@0 77 % maximum posterior prob.
wolffd@0 78 class = class';
wolffd@0 79 for m = 1:class_num,
wolffd@0 80 thisX=[]; thisY=[];
wolffd@0 81 thisX = X(class == m);
wolffd@0 82 thisY = Y(class == m);
wolffd@0 83 if class_num<=6,
wolffd@0 84 str_col=[];
wolffd@0 85 str_col=['d', colors(m,:)];
wolffd@0 86 h=plot(thisX, thisY, str_col);
wolffd@0 87 else
wolffd@0 88 h = plot(thisX, thisY, 'd',...
wolffd@0 89 'MarkerEdgeColor',colors(m,:),...
wolffd@0 90 'MarkerFaceColor','w');
wolffd@0 91 end
wolffd@0 92 set(h, 'MarkerSize', 4);
wolffd@0 93 end
wolffd@0 94 title('Training set and Decision Boundaries (0/1 loss)')
wolffd@0 95 hold off
wolffd@0 96
wolffd@0 97 % Plot of test_set --------------------------------------------------------------------------
wolffd@0 98 if nargin==4,
wolffd@0 99 subplot(1,2,2);
wolffd@0 100 axis([mn_x-1 mx_x+1 mn_y-1 mx_y+1]);
wolffd@0 101 set(gca, 'Box', 'on');
wolffd@0 102 hold on
wolffd@0 103 if size(test_data,2)==3, % we know the classification of the test set examples
wolffd@0 104 c_max_test = max(test_data(:,3));
wolffd@0 105 for m=1:c_max_test,
wolffd@0 106 app_x=test_data(:,1);
wolffd@0 107 app_y=test_data(:,2);
wolffd@0 108 thisX=app_x(test_data(:,3)==m);
wolffd@0 109 thisY=app_y(test_data(:,3)==m);
wolffd@0 110 if class_num<=6,
wolffd@0 111 str_col=[];
wolffd@0 112 str_col=['o', colors(m,:)];
wolffd@0 113 plot(thisX, thisY, str_col, 'MarkerSize', ms);
wolffd@0 114 else
wolffd@0 115 plot(thisX, thisY, 'o',...
wolffd@0 116 'LineWidth', 1,...
wolffd@0 117 'MarkerEdgeColor', colors(m,:),...
wolffd@0 118 'MarkerSize',ms);
wolffd@0 119 end
wolffd@0 120 end
wolffd@0 121 else
wolffd@0 122 plot(test_data(:,1), test_data(:,2), 'ko',...
wolffd@0 123 'MarkerSize', ms);
wolffd@0 124 end
wolffd@0 125 for m = 1:class_num,
wolffd@0 126 thisX=[]; thisY=[];
wolffd@0 127 thisX = X(class == m);
wolffd@0 128 thisY = Y(class == m);
wolffd@0 129 if class_num<=6,
wolffd@0 130 str_col=[];
wolffd@0 131 str_col=['d', colors(m,:)];
wolffd@0 132 h=plot(thisX, thisY, str_col);
wolffd@0 133 else
wolffd@0 134 h = plot(thisX, thisY, 'd',...
wolffd@0 135 'MarkerEdgeColor', colors(m,:),...
wolffd@0 136 'MarkerFaceColor','w');
wolffd@0 137 end
wolffd@0 138 set(h, 'MarkerSize', 4);
wolffd@0 139 end
wolffd@0 140 title('Test set and Decision Boundaries (0/1 loss)')
wolffd@0 141 hold off
wolffd@0 142 end