
II. File Formats

Outline
 A. Summary of File Types
 B. Data file include in this package
  
 A. Summary of File Types
    The MLP,Piecewise Linear Net,Functional link net and the 
    Unsupervised learning programs typically have the following file type
    and file formats associated with them. 
    a.Training file for Mapping :The training file
      for mapping has the following file format:It has
      inputs followed by the outputs and the training patterns
      as its number of rows.For example training file Twod.tra has
      8 inputs,7 outputs and 1768 patterns.
      8         7           
    b.Training file for Classification :The training file
      for classification has the following file format.It has
      inputs followed by the outputs and the number of rows 
      is its training patterns. The output has coded format or uncoded format.
    c.Testing file for Mapping:The testing file for mapping has the following
      file format.It has inputs followed by the outputs in one column and the
      number of rows is its training patterns.For example testing file Twod.tra
      has 8 inputs,7 outputs and 1768 patterns.
      8         7
    d.Testing file for Classification:The testing file
      for classification has the following file format.It has
      inputs followed by the outputs and the number of rows 
      is its training patterns. The output has coded format or uncoded format.
    e.Processing file for Mapping:The processing file
      for classification has the following file format.It has
      inputs followed by the outputs and the number of rows 
      is its training patterns.For example processing file Twod.tra has
      8 inputs,7 outputs and 1768 patterns.
      8         7
    f.Processing file for classification:The processing file
      for classification has the following file format.It has
      inputs followed by the outputs and the number of rows 
      is its training patterns. The output has coded format or uncoded format.
    g.Weights file for MLP:The trained weight file for the MLP pruning,
      testing and processing is of the following file format.For example the 
      weights file MLP.wts has the following format,
      twod.tra                            !     training file name
      8                                   !     number of inputs
      7                                   !     number of outputs
      10                                  !     number of hidden units
      The Training _alogorithm is OWO_HWO !     training algorithm name
      followed by N(number of input) rows with Nout(number of output) columns
      of weights connecting each input to each output. followed by the output's threshold.
      followed by Nh(number of hidden layers) rows with Nout columns 
      of weights connecting each hidden layer to each output.
      followed by Nh rows with N+1(including a threshold of energy 1) columns
      of weights connecting each input to each hidden unit.
    h.Weights file for PLN:The trained weight file for the PLN  testing and
      processing is of the following format.For example the weights file PLN.wts
      has the following format,
      twod.tra                            !     training file name
      8                                   !     number of inputs
      7                                   !     number of outputs
      10                                   !     number of clusters
      followed by Nout rows with N+1 columns of global linear weights,
      followed by Nout rows with N columns of distance measure for each input,
      followed by Nout rows with N columns of mean of the inputs,
      followed by Nout rows with N columns of variance of the inputs,
      followed by Nc(number of clusters) blocks with each block has Nout rows with N+1 
      columns of modular weights,
      followed by Nc(number of clusters) rows with N columns of cluster centers,
      followed by Nc columns of outlier distance.
    i.Weights file for FLN:The trained weight file for the FLN  testing and
      processing is of the following format.For the functional link net, this file
      contains the network degree P(usually an integer between 1 and 3), the number
      of network inputs N and the number of outputs, and the dimension of the 
      multinomial vector,which is L = (N+P)!/(N!P!).For example the weights file
      FLN.wts has the following format,
      twod.tra                            !     training file name
      8                                   !     number of inputs
      7                                   !     number of outputs
      2                                   !     degree of the network
?     followed by Nout rows with L columns of weights.
    j.Clusters center file for Unsupervised Learning: The trainined weight file for
      unsupervised learning such as Kmean and SOM has the following
      file format.For example the weights file clusters.wts has the following format,
      twod.tra                                   !     training file name
      15                                         !     total number of columns(inputs+outputs)
      8                                          !     number of columns to be clustered
      20                                         !     number of clusters
      The Training _alogorithm is SOM_Clustering !     training algorithm name
      followed by  Nc(number of clusters) rows with N columns of cluster weights.
    k.Weights file for KLT matrix: The calculated weight file for
      KLT matrix for compression has the following file format.
      For example the weights file KLTmatrix.klt has the following format,
      twod.tra                                   !     training file name
      15                                         !     total number of columns(inputs+outputs)
      1                                          !     starting column # for compression
      20                                         !     ending column # for compression
      followed by  N rows with N columns of KLT matrix.  
   
    l.The result file, which has the extension result.res. This ASCII file
      stores training results, such as the minimum mean-square error 
      attained during training,testing and processing.
 
 B. Data Files Included With This Package
    The twod.tra data file has 1,768 patterns with 8 inputs and 7 desired 
    outputs. The data comes from a remote sensing problem.


