pnn_vc.m

最新的模式识别分类工具箱,希望对朋友们有用!

% Learns classifier and classifies test set
% using a Probabilistic Neural Network
% Usage
%      [trainError, testError, estTrainLabels, estTestLabels] = ...
%           PNN_VC(trainFeatures, trainLabels,Nmu ,testFeatures, testLabels)
% where
%
% Inputs:
% 	trainFeatures  - the training set vectors, one vector per column
%	trainLabels    - the labels of the above
%	sigma          - Gaussian width
%       testFeatures   - test set, one column per vector
%       testLabels     - labels  for test set%
% Outputs
%	trainError     - the error rate on the training set (one entry per
%	                   class + total error)
%	testError      - the error rate on the test set (one entry per class
%	                  + total error)
%       estTrainLabels - the labels produced by the algorithm for the
%                          training samples
%       estTestLabels - the labels produced by the algorithm for the
%                          test samples

function [trainError, testError, estTrainLabels, estTestLabels] = ...
    PNN_VC(trainFeatures, trainLabels,sigma ,testFeatures, testLabels)

hm = findobj('Tag', 'Messages'); 
fprintf('Probabilistic Neural Network: Training\n');
if (isempty(hm)==0)
  s = sprintf('Probabilistic Neural Network: Training');
  set(hm,'String',s);
  refresh;
end


[Dim, Nsam]       = size(trainFeatures);
normFeatures    = trainFeatures;
normFeatures(Dim+1,:) = ones(1,Nsam);
[Nclasses, classes]  = find_classes([trainLabels(:);testLabels(:)]); % Number of classes in labels


%  Build the classifier
%  Normalizes the training features

normFeatures  = normFeatures ./ (sqrt(sum(normFeatures'.^2))'*ones(1,Nsam));

%
%

for cl=1:Nclasses
  evalstr =sprintf('ind%d = find(trainLabels == classes(cl));', cl);
  eval (evalstr);
  evalstr =sprintf('ClassFeatures%d = normFeatures(:,ind%d);',cl,cl);
  eval (evalstr);  
  evalstr =sprintf('Onen%d =ones(size(ind%d));', cl,cl);
  eval (evalstr);
  evalstr =sprintf('Num(%d) = length(ind%d);',cl, cl);
  eval(evalstr);
end
disp(Num);

fprintf('Probabilistic Neural Network: Computing Training Set Error Rate\n');
if (isempty(hm)==0)
  s = sprintf('Probabilistic Neural Network: Computing Training Set Error Rates');
  set(hm,'String',s);
  pause(.1);
end


classVote=[];
for sam=1:Nsam,
  thisSam =[trainFeatures(:,sam);1];
  %thisSam =trainFeatures(:,sam);
  for cl = 1:Nclasses,

    evalstr = ...
	sprintf('innerprod = (thisSam * Onen%d).* ClassFeatures%d;',cl,cl);
    eval(evalstr);
    sumval    = (sum(innerprod) - 1)/(sigma * sigma);
    classVote(cl) = sum(exp(sumval))/Num(cl);
  end
  [foo,bestClass] = max(classVote);
  estTrainLabels(sam)= classes(bestClass);
end

fprintf('Probabilistic Neural Network: Computing Test Set Error Rate\n');
if (isempty(hm)==0)
  s = sprintf('Probabilistic Neural Network: Computing Test Set Error Rates');
  set(hm,'String',s);
  pause(.1);
end


[Dim, Nsamt]       = size(testFeatures);
classVote=[];
for sam=1:Nsamt,
  thisSam =[testFeatures(:,sam);1];
  %thisSam =testFeatures(:,sam);
  for cl = 1:Nclasses,
    evalstr = ...
	sprintf('innerprod = (thisSam * Onen%d).* ClassFeatures%d;',cl,cl);
    eval(evalstr);
    sumval        = (sum(innerprod) - 1)/(sigma * sigma);
    classVote(cl) = sum(exp(sumval))/Num(cl);
  end
  [foo,bestClass] = max(classVote);
  estTestLabels(sam)= classes(bestClass);
end

    
trainError = computeError( classes, trainLabels, estTrainLabels);
testError = computeError ( classes, testLabels, estTestLabels);