prtClassRvmSequential  Relevance vector machine classifier using sequential training
  
    CLASSIFIER = prtClassRvmSequential returns a relevance vector
    machine classifier based using sequential training.
 
    CLASSIFIER = prtClassRvmSequential(PROPERTY1, VALUE1, ...)
    constructs a prtClassRvmSequential object CLASSIFIER with properties as
    specified by PROPERTY/VALUE pairs.
 
     A prtClassRvmSequential object inherits all properties from the abstract class
     prtClass. In addition is has the following properties:
 
     kernels                - A cell array of prtKernel objects specifying
                              the kernels to use
     verbosePlot            - Flag indicating whether or not to plot during
                              training
     verboseText            - Flag indicating whether or not to output
                              verbose updates during training
     learningMaxIterations  - The maximum number of iterations
 
     A prtClassRvmSequential also has the following read-only properties:
 
     learningConverged  - Flag indicating if the training converged
     beta               - The regression weights, estimated during training
     sparseBeta         - The sparse regression weights, estimated during
                          training
     sparseKernels      - The sparse regression kernels, estimated during
                          training
 
    For more information on the algorithm and the above properties, see
    the following reference:
 
    Tipping, M. E. and A. C. Faul (2003). Fast marginal likelihood
    maximisation for sparse Bayesian models. In C. M. Bishop and B. J.
    Frey (Eds.), Proceedings of the Ninth International Workshop on
    Artificial Intelligence and Statistics, Key West, FL, Jan 3-6.
 
    prtClassRvmSequential is most useful for datasets with a large
    number of observations for which the gram matrix can not be held in
    memory. The sequential RVM training algorithm is capable of
    operating by generating necessary portions of the gram matrix when
    needed. The size of the generated portion of the gram matrix is
    determined by the property, largestNumberOfGramColumns. Sequential
    RVM training will attempt to generate portions of the gram matrix
    that are TraingData.nObservations x largesNumberofGramColums in
    size. If the entire gram matrix is this size or smaller it need
    only be generated once. Therefore if the entire gram matrix can be
    stored in memory, training is much faster. For quickest operation,
    largestNumberOfGramColumns should be set as large as possible
    without exceeding RAM limitations.
 
     Example:
 
     TestDataSet = prtDataGenUnimodal;      % Create some test and
     TrainingDataSet = prtDataGenUnimodal;  % training data classifier
     classifier = prtClassRvmSequential('verbosePlot',true); % Create a classifier
     classifier = classifier.train(TrainingDataSet);    % Train
     classified = run(classifier, TestDataSet);         % Test
     % Plot
     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');