import_ascii_data_snn.m

来自「神经网络的工具箱, 神经网络的工具箱,」· M 代码 · 共 81 行

function data = import_ascii_data_snn(filename, input, output, weight, missing)
%IMPORT_ASCII_DATA_SNN import data from ascii file in wcfdata_struct.
%
%  Syntax
%
%   wcfdata_struct = import_ascii_data_snn(filename, input_columns, ...
%                    target_columns) 
%   wcfdata_struct = import_ascii_data_snn(filename, input_columns, ...
%                    target_columns, weight_column) 
%   wcfdata_struct = import_ascii_data_snn(filename, input_columns, ...
%                    target_columns, weight_column, missing_value_indicator) 
%   wcfdata_struct = import_ascii_data_snn(filename, input_columns, ...
%                    target_columns, [], missing_value_indicator) 
% 
%  Description
%
%   IMPORT_ASCII_DATA_SNN takes
%     filename       - The name of the ascii file. For each pattern this
%                      file must contain a row with space seperated values 
%                      for inputs, targets and pattern weights (if any).
%     input_columns  - 1 x #inputs matrix of column numbers for the inputs.
%     target_columns - 1 x #targets matrix of column number for the targets. 
%     weight_column  - column number for pattern weights. [] for no pattern 
%                      weights (default).
%     missing_value_indicator - a value indicating the target is not
%                               available and/or should not be used in 
%                               training (default NaN).
%   and returns a wcfdata_struct which can be used by WCF_SNN.
%
%  Examples
%
%   Given the file 'file.asc':
%   ------file.asc(begin)------------------------------------------------
%   213 34 3123 324 1314 654321 1 
%   312 53 4122 264 654321 4341 0.7 
%   621 24 3223 324 654321 5312 0.7
%   ------file.asc(end)--------------------------------------------------
%
%   the command:
%   
%   wcfdata = import_ascii_data_snn('file.asc', [1:3 5], [4 6], 7, 654321)
%
%   will return a wcfdata_struct representing 3 patterns, with
%   inputs from column 1,2,3 and 5, targets from columns 4 and 6 and 
%   pattern weights from column 7.
%   In training, pattern 1 will be used to train only on the first
%   target, while pattern 2 and 3 will be used only to train on the
%   second target. 
%
%  See also
%
%   WCFDATA_STRUCT_SNN, TRAIN_SNN
%

if (nargin <5)
   missing = NaN;
end

S = load_ascii_snn(filename);

data.P = S(:, input)';
data.T = S(:, output)';
if (nargin>3)
   data.gmu = S(:, weight)';
end
if (isnan(missing)) 
   missing_ind = find(isnan(data.T));
   if missing_ind
      data.useT = ones(size(data.T));
      data.useT(missing_ind) = 0;
   end
else
   missing_ind = find(missing == data.T);
   if missing_ind
      data.useT = ones(size(data.T));
      data.useT(missing_ind) = 0;
   end
end