notes:
 
1.  The demo dataset and plots were constructed from data contained in the
    PROBEN1 benchmark collection. The benchmark set is available for anonymous
    FTP at ftp.cs.cmu.edu /afs/cs/project/connect/bench/contrib/prechelt.
    See 
        "Proben1 -- A Set of Neural Network Benchmark Problems and
        Benchmarking Rules,"
        Lutz Prechelt,
        Fakultat fur Informatik,
        Universitat Karlsruke, 1994.

2.  Code information:

    grnnR assumes "topological" organization with: 
    -column-major observations (pattern matrix, P)
    -each row represents a domain variable
    -column-major range targets and grnnR estimates

    TERMINOLOGY
  
    domain -> input space -> observations -> pattern vectors
    grnnR -> nnet -> neurons -> center vectors -> weight vectors
    nnet -> neurons -> Gaussian kernels -> spread -> SIGMA
    range -> output space -> target vectors (reference, estimated)

    SYMBOL LIST

    N: Nr observations
       scalar
    n: Nr neurons
       scalar
    d: input space dimension
       scalar
    r: output space dimension
       scalar
    p: pattern vector
       vector, d x 1
    P: matrix of pattern vectors
       matrix, d x N
    t: target vector
       vector, r x 1
    T: matrix of target vectors
       matrix, r x N
    t_hat: estimated target vector
       vector, r x 1
    T_hat: matrix of estimated target vectors
       matrix, r x N
    c: center vector
       vector, d x 1
    C: matrix of center vectors
       matrix, d x n
    w: weight vector
       vector, r x 1
    W: matrix of weight vectors
       matrix, r x n
    s: output of summation neuron
       scalar
    SIGMA: Gaussian spread value
       scalar
    err: error vector
       vector, r x 1
    ERR: matrix of error vectors
       matrix, r x N
    X_mc: mean-centered matrix
       matrix, dim of X
    mn_Vect: vector of row (variable) means
       vector, d x 1
    X_var: variance-scaled matrix
       matrix, dim of X
    var_Vect: vector of variable (row) variances
       vector, d x 1

    FUNCTIONS

    1. grnn_Train
    2. grnn
    3. compute_SIGMA
    4. test_Stats
    5. plot_Targets
    6. plot_Box
    7. mean_center_X
    8. var_scale_X

3.  Example/demo:
    A demo script is provided which loads a sample dataset (P_trn, T_trn, P_tst,
    and T_tst), synthesizes the grnn, then performs batch testing. An error
    plot is generated to the terminal and saved to file grnnPlot.eps. To load
    the demo, type "demo(demo_TRAIN)" followed by "demo(demo_TEST)". See the
    script files demo_TRAIN and demo_TEST in the demo subdirectory for a usage
    example of the grnnR package.
