inverse_error_snn.m

神经网络的工具箱, 神经网络的工具箱,

function [yl, yu] = inverse_error_snn(y_av, e_max, errf)
%INVERSE_ERROR_SNN inverse of errorfunction.
%
% Syntax
%
%  [yl, yu] = inverse_error_snn(y_av, err, error_f)
%
% Description
%
%  INVERSE_ERROR_SNN takes
%   y_av     - error is around y_av
%   err      - crossentropy error
%   error_f  - name of the error function to be inversed
%  and returns
%   yl       - lowest y for which error_f(y, y_av) = err
%   yu       - highest y for which error_f(y, y_av) = err
%  
%  supported error functions:
%     SE_SNN, RELERR_SNN, LOGLIKELIHOOD_SNN, CROSSLOGISTIC_SNN,
%     CROSSENTROPY_SNN
%


[N, MU] = size(y_av);

%#function se_snn invse_snn
%#function relerr_snn invrelerr_snn
%#function loglikelihood_snn invloglikelihood_snn
%#function crosslogistic_snn invcrosslogistic_snn
%#function crossentropy_snn invcrossentropy_snn
yl = zeros(N,MU);
yu = zeros(N,MU);
if (isstr(errf))
   [yl, yu] = feval(feval(errf, 'inv'), y_av, e_max);   
elseif (iscell(errf))
   for i = 1:N
       [yl(i,:), yu(i,:)] = ...
           feval(feval(errf{i,1}, 'inv'), y_av(i,:), e_max(i,:));
   end
else
   error('Incorrect format for output errorfunction');
end