MATLAB File Help: prtClassAdaBoost
prtClassAdaBoost
 prtClassAdaBoost AdaBoost classifier
 
     CLASSIFIER = prtClassAdaBoost returns a AdaBoost classifier
 
     CLASSIFIER = prtClassAdaBoost(PROPERTY1, VALUE1, ...) constructs a
     prtClassAdaBoost object CLASSIFIER with properties as specified by
     PROPERTY/VALUE pairs.
 
     A prtClassAdaBoost object inherits all properties from the abstract
     class prtClass. In addition is has the following properties:
 
     baseClassifier     - the prtClass object that forms the "weak" or
                          "Base" classifier for the AdaBoost.  This
                          classifier is iteratively trained on each of
                          the features in succession.
     maxIters            - Number of iterations to run (maximum number of weak
                          classifiers to train)
     deltaPeThreshold (.05) - Specify the minimum error distance (from
                         0.5) allowed.  Smaller values result in more
                         complicated adaBoost classifiers, larger
                         values in less complicated classifiers.
     
     downSampleBootstrap (false) - Specify whether (and how many)
                            bootstrap-by class bootstrap samples to
                            take at each iteration.  False or 0 uses
                            the default number of bootstrap samples.
                            Any other integer specifies the number of
                            samples to use from
                            ds.bootstrapDataByClass.
 
     AdaBoost is a meta algorithm for training ensembles of weak
     classifiers on different sub-sets of the complete data set, with
     later classifiers trained to focus on data points that were
     mis-classified in earlier iterations.  A complete description of the
     algorithm for AdaBoost is beyond the scope of this help entry, but
     more information AdaBoost can be found at the following URL:
 
     http://en.wikipedia.org/wiki/AdaBoost
 
     The PRT AdaBoost uses individual features to constitute each weak
     learner in each iteration.  At each iteration, the feature
     corresponding to the best weak learner operating on the current
     weighted data distribution is used in the aggregate classifier.
 
     A prtClassAdaBoost object inherits the TRAIN, RUN, CROSSVALIDATE and
     KFOLDS methods from prtAction. It also inherits the PLOT
     method from prtClass.
 
     Example:
 
     TestDataSet = prtDataGenUnimodal;       % Create some test and
     TrainingDataSet = prtDataGenUnimodal;   % training data
     classifier = prtClassAdaBoost;                     % Create a classifier
     classifier = classifier.train(TrainingDataSet);    % Train
     classified = run(classifier, TestDataSet);         % Test
     subplot(2,1,1);
     classifier.plot;
     subplot(2,1,2);
     [pf,pd] = prtScoreRoc(classified,TestDataSet);
     h = plot(pf,pd,'linewidth',3);
     title('ROC'); xlabel('Pf'); ylabel('Pd');
See also
Class Details
Superclasses prtClass
Sealed false
Construct on load false
Constructor Summary
prtClassAdaBoost AdaBoost classifier 
Property Summary
baseClassifier The weak classifier 
dataSet The training prtDataSet, only stored if verboseStorage is true.  
dataSetSummary Structure that summarizes prtDataSet. 
deltaPeThreshold  
downSampleBootstrap  
internalDecider Optional prtDecider object for making decisions 
isCrossValidateValid True 
isNativeMary False 
isSupervised True 
isTrained Indicates if prtAction object has been trained. 
maxIters Max number of iterations 
name AdaBoost 
nameAbbreviation AdaBoost 
showProgressBar  
twoClassParadigm Whether the classifier retures one output (binary) or two outputs (m-ary) when there are only two unique class labels 
userData User specified data 
verboseStorage Specifies whether or not to store the training prtDataset. 
Method Summary
  crossValidate Cross validate prtAction using prtDataSet and cross validation keys. 
Static   fastRoc  
  get get the object properties 
  kfolds Perform K-folds cross-validation of prtAction 
  optimize Optimize action parameter by exhaustive function maximization. 
  plot Plot the output confidence of a prtClass object 
  run Run a prtAction object on a prtDataSet object. 
  set set the object properties 
  train Train a prtAction object using training a prtDataSet object.