fuz_id.m

The neuro-fuzzy software for identification and data analysis has been implemented in the MATLAB lan

function [centers,bases,par,bias]=fuz_id(comp,training_set)

%FUZ_ID		Identification of the parameters of a Fuzzy Model
%
%  [centers,bases,par,bias]=fuz_id(comp,training_set)
%
%	centers[no_rules,no_var]
%	bases[no_rules,no_var]
%	par[no_rules,no_par]            where no_par=1 if model='constant'
%				     and no_par=no_var+1 if model='linear'
%	bias[1,1]
%	
%	comp[1,1]	   is the complexity of the model being identified
%	training_set[no_data,no_var+1]	  the first no_var columns are the
%				        inputs, the last one is the output
%
%	The following global variable are required:                     
%		memb_fun:	'gaussian'/'triangular'                
% 		arch:		'weigthed'/'comb'                      
%		format_out:	'linear'/'constant'                       
%		init: 		'k_mean'/'hfc'
%		fid_h

%______________________________________________________10 April 1996____

global memb_fun
global arch
global format_out
global init

global fid_h


%Initialization:

[centers,bases]=eval([init '(comp,training_set)']);


%Optimization:

termination=0;
old_SSE=1e+10;
k=0.005;
[no_data,c]=size(training_set);
in_trn=training_set(:,1:c-1);
out_trn=training_set(:,c);


while (termination <2)
				
	% INVERSION to determine the parameters of the CONSEQUENTS

		if strcmp(format_out,'constant')
			[SSE,new_par,bias]=inv_co(in_trn,centers,bases,out_trn);
		end

		if strcmp(format_out,'linear')
			[SSE,new_par,bias]=inv_li(in_trn,centers,bases,out_trn);
		end
		
		

		if (SSE/old_SSE >1-k)
			termination=termination+1;
		end
		
		if isnan(SSE)
			break;
		end

		par=new_par;

		
fprintf(fid_h,'\t\t\t\tConsequents optimization. SSE=%g\n',SSE);

		old_SSE=SSE;
			
	% LEVENBERG-MARQUARDT to determine the CENTERS and the BASES

		[centers,bases]=leve_mao(in_trn,centers,bases,par,bias,out_trn);

end