function [Inter,Knots,numb_cells,n_fun_over,n_fun,weights,disp_matrix]=create_CMAC(min,max,num_knot,gener,disp);

%	[Knots,numb_cells,n_fun_over,n_fun,weights,disp_matrix]=create_CMAC(min,max,num_knot,gener,disp)
%
%	This function creates a CMAC neural network assuming
%	that the minimum value of the input vector is stored in vector
%	min, the maximum value of the input vector is stored in vector max,
%  the number of internal knots is stored in vector num_knots,
% 	and the generalization parameter is gener. Disp is the displacement vector
%	for the first overlay
%	It is assumed that the interior knots are equidistantly distributed 
%	and that binary CMACs, with active functions with the value of
%	1/gener are employed.
%
%	Output parameters:
%
%	Knots - matrix where the knots in ecah input dimension are stored 
%	by row
%
%	numb_cells - total number of cells in the network
%
%	n_fun_over - number of basis functions per overlay
%
%	n_fun - number of basis functions for the newtork
%
%	weights - vector of weights in the network
%
ndim=length(min);

%	compute the displacement matrix
disp_matrix=disp;
for i=1:gener-1
   for j=1:ndim
      disp_matrix(j,i+1)=mod((i+1)*disp_matrix(j,1),gener);
      if disp_matrix(j,i+1)==0
         disp_matrix(j,i+1)=gener;
      end   
   end   
end   
   
for i=1:ndim
   delta_x=max(i)-min(i);
   delta=delta_x/(num_knot(i)+1);
   Inter(i,1:num_knot(i)+2)=[min(i):delta: max(i)];
   Knots(i,1:num_knot(i))=Inter(i,2:num_knot(i)+1);
   for j=1:gener
      sup(i,j)=ceil((num_knot(i)+1-disp_matrix(i,j))/gener)+1;
   end   
end   

n_fun_over=prod(sup,1);
n_fun=sum(n_fun_over);
numb_cells=prod(num_knot+1);
weights=zeros(n_fun,1);