som_set

Purpose

 Create and set values for fields of SOM Toolbox structs (except
 visualization struct). Can also be used to check the validity of structs.

Syntax

  sMap   = som_set('som_map');
  sData  = som_set(sData);
  sNorm  = som_set(...,'field',contents,...);
  [sTopol,ok]      = som_set(sTopol,...);
  [sTrain,ok,msgs] = som_set('som_train',...);

Description

 The function is used to create and set values for fields of SOM
 Toolbox structs, except visualization structs. The given values are
 first checked for validity, and if they are not valid, an error
 message is returned. The function can also be used to check the
 validity of all the fields of the struct by supplying a struct as
 the first and only argument.

 NOTE: Using SOM_SET to create structures does _not_ guarantee that the
 structs are valid (try e.g. sM = som_set('som_map'); som_set(sM)). The
 initial values that the function gives to the fields of the structs are
 typically invalid. It is recommended that when creating map or data 
 structs, the corresponding functions SOM_MAP_STRUCT and SOM_DATA_STRUCT 
 are used instead of SOM_SET. However, when giving values for the fields, 
 SOM_SET tries to guarantee that the values are valid.

 If a string is given as the first argument, the corresponding 
 structure is first created and the field-content pairs are then
 applied to it. 

 There can be arbitrarily many field-contents pairs. The pairs
 are processed sequentially one pair at a time. For each pair,
 the validity of the contents is checked and the corresponding 
 items in the returned 'ok'-vector and 'msgs'-cellstring are set.
 - if the contents is ok, the status is set to 1 and message to ''
 - if the contents is suspicious, status is set to 1, but a
   message is produced
 - if the contents is invalid, status is set to 0 and an error
   message is produced. The contents are _not_ given to the field.
 If there is only one output argument, the status and messages
 for each pair are printed to standard output.

 The different field-contents pairs have no effect on each other.
 If a field is given a value multiple times, the last valid one 
 stays in effect.

 In some cases, the order of the given fields is significant.
 For example in the case of 'som_map', the validity of some fields, 
 like '.comp_names', depends on the input space dimension, which is
 checked from the '.data' field (dim = size(sD.data,2) to be specific).
 Therefore, the '.data' field (or '.codebook' field in case of map 
 struct) should always be given a value first. Below is a list of 
 this kind of dependancies:

 som_map:   'comp_names', 'comp_norm', 'msize', 'topol.msize',
            'labels' and 'mask' depend on 'codebook'
            new value for 'codebook' should have equal size to the old
            one (unless the old one was empty)
 som_data:  'comp_names' and 'comp_norm' depend on 'data'
            new value for 'data' should have equal dimension (size(data,2))
            as the old one (unless the old one was empty)

Known bugs

 Checking the values given to som_grid struct has not been
 implemented. Use SOM_GRID function to give the values.

Required input arguments

  sS          The struct.
     (struct) A SOM Toolbox struct.
     (string) Identifier of a SOM Toolbox struct: 'som_map', 
              'som_data', 'som_topol', 'som_norm' or 'som_train'

Optional input arguments

  field     (string) Field identifier string (see below).
  contents  (varies) Value for the field (see below).

  Below is the list of valid field identifiers for the different 
  SOM Toolbox structs. 

  'som_map' (map struct)
    'codebook'    : matrix, size [munits, dim] 
    'labels'      : cell array of strings, 
                    size [munits, maximum_number_of_labels]
    'topol'       : topology struct (prod(topol.msize)=munits)
    'mask'        : vector, size [dim, 1]
    'neigh'       : string ('gaussian' or 'cutgauss' or 'bubble' or 'ep')
    'trainhist'   : struct array of train structs
    'name'        : string
    'comp_names'  : cellstr, size [dim, 1], e.g. {'c1','c2','c3'}
    'comp_norm'   : cell array, size [dim, 1], of cell arrays 
                    of normalization structs
    Also the following can be used (although they are fields
    of the topology struct)
    'msize'       : vector (prod(msize)=munits)
    'lattice'     : string ('rect' or 'hexa')
    'shape'       : string ('sheet' or 'cyl' or 'toroid')

  'som_data' (data struct)
    'data'        : matrix, size [dlen, dim]
    'name'        : string
    'labels'      : cell array of strings, 
                    size [dlen, m]
    'comp_names'  : cellstr, size [dim, 1], e.g. {'c1','c2','c3'}
    'comp_norm'   : cell array, size [dim, 1], of cell arrays 
                    of normalization structs
    'label_names' : cellstr, size [m, 1]

 'som_topol' (topology struct)
    'msize'       : vector
    'lattice'     : string ('rect' or 'hexa')
    'shape'       : string ('sheet' or 'cyl' or 'toroid')

 'som_norm' (normalization struct)
    'method'      : string
    'params'      : varies
    'status'      : string ('done' or 'undone' or 'uninit')

 'som_train' (train struct)
    'algorithm'   : string ('seq' or 'batch' or 'lininit' or 'randinit')
    'data_name'   : string
    'mask'        : vector, size [dim, 1]
    'neigh'       : string ('gaussian' or 'cutgauss' or 'bubble' or 'ep')
    'radius_ini'  : scalar
    'radius_fin'  : scalar
    'alpha_ini'   : scalar
    'alpha_type'  : string ('linear' or 'inv' or 'power')
    'trainlen'    : scalar
    'time'        : string

 'som_grid' (grid struct) : checking the values has not been implemented yet!
    'neigh'       : string ('rect' or 'hexa') or 
                    (sparce) matrix, size munits x munits
    'shape'       : string ('sheet' or 'cyl' or 'toroid')
    'msize'       : vector, size 1x2
    'coord'       : matrix, size munits x 2 or munits x 3
    'line'        : string (linespec, e.g. '-', or 'none')
    'linecolor'   : RGB triple or string (colorspec, e.g. 'k') or 
                    munits x munits x 3 (sparce) matrix or cell
                    array of RGB triples 
    'linewidth'   : scalar or munits x munits (sparce) matrix
    'marker'      : string (markerspec, e.g. 'o', or 'none') or 
                    munits x 1 cell or char array of these
    'markersize'  : scalar or munits x 1 vector
    'markercolor' : RGB triple or string (colorspec, e.g. 'k')
    'surf'        : [], munits x 1 or munits x 3 matrix of RGB triples
    'label'       : [] or munits x 1 char array or 
                    munits x l cell array of strings 
    'labelcolor'  : RGB triple or string (colorspec, e.g. 'g' or 'none')
    'labelsize'   : scalar

Output arguments

  sS    (struct)  the created / updated struct
  ok    (vector)  length = number of field-contents pairs, gives
                  validity status for each pair (0=invalid, 1 otherwise)
  msgs  (cellstr) length = number of field-contents pairs, gives
                  error/warning message for each pair ('' if ok)

Examples

 To create a struct:
  sM  = som_set('som_map');
  sD  = som_set('som_data');
  sTo = som_set('som_topol');
  sTr = som_set('som_train');
  sN  = som_set('som_norm');
  sG  = som_set('som_grid');

 To check the the contents of a struct: 
  som_set(sS);
  [dummy,ok]      = som_set(sS);
  [dummy,ok,msgs] = som_set(sS);

 To give values to fields: 
  sTo = som_set(sTo,'msize',[10 10],'lattice','hexa','shape','toroid');
  sM  = som_set('som_map','codebook',rand(100,4),'topol',sTo);

See also