LogitBooster

 function [lb] = LogitBooster(rg)
    constructor of the LogitBooster class that inherits from the classifier
    class. the LogitBooster implements the LogitBoost boosting algorithm to
    build a strong (boosted) classifier from several weak classifiers.

    Inputs:
       rg: the regressor needed to for logitboost
       numClasses: number of classes

0001 function [lb] = LogitBooster(rg, numClasses)
0002 % function [lb] = LogitBooster(rg)
0003 %    constructor of the LogitBooster class that inherits from the classifier
0004 %    class. the LogitBooster implements the LogitBoost boosting algorithm to
0005 %    build a strong (boosted) classifier from several weak classifiers.
0006 %
0007 %    Inputs:
0008 %       rg: the regressor needed to for logitboost
0009 %       numClasses: number of classes
0010 
0011 
0012 % parameters needed for training
0013 % the error bound after reaching which the classifier stops learning,
0014 % used only when the nStages argument to learn is Inf
0015 lb.errBound = 0.001;
0016 
0017 % paramters needed to define the classifier
0018 if nargin == 0
0019     lb.regressor = Nan;
0020 else
0021     lb.regressor = rg;
0022 end
0023 
0024 lb.nStages = 0;
0025 
0026 % 1. number of classes
0027 lb.numClasses = numClasses;
0028 
0029 % 2. example weights of the last iteration of LogitBoost learning
0030 % algorithm. this is saved after training the classifier to be used
0031 % later if we want to increase the number of stages later on
0032 lb.lastExWeights = [];
0033 
0034 % 3. the threshold whereby the classifier can distinguish between the
0035 % positive and negative examples
0036 lb.thresh = NaN;
0037 
0038 % 4. detection rate after training
0039 lb.detectionRate = NaN;
0040 
0041 % 5. F matrix (J,n) (j number of classes, n is the number of features)
0042 lb.F = NaN;
0043 
0044 % 6. P matrix (m, n) (m number of instances , n is the number of features)
0045 lb.P = NaN;
0046 
0047 lb = class(lb, 'LogitBooster', Classifier(numClasses));