auto_elevator_design.m

其中提到遺傳學的程式碼與應用提供給次淚相向的研究者參考下載

% auto_elevator_design.m
% Use FFNN to design a special ELEVATOR. 
% Refer to Demuth's Neural Network Design textbook.
% This is the example on P.6-147 of Chou's textbook.

% PenChen Chou, Aug. 20, 2001
% Revised on    Apl. 21, 2002
% ChiZen  May , Aug. 20, 2001

global net P T Emin MUL_factor MIN_offset

% Use 5-2 FFNN, given input and target patterns
P=[   0  0  0 1  1 -1   1 -1 -1 1 -1  -1 -1  1  1  1 1 -1
      0  0  0 1 -1  1  -1  1 -1 1 -1   1 -1 -1  1 -1 1 -1
      1  0  0 1  0  0   0  1  1 1  0   0  0  0  0  1 0  0
      0  1  0 0  1  0   0  0  0 1  1   1  0  0  0  0 1  0
      0  0  1 0  0  1   0  0  0 0  0   0  1  1  1  0 0  1];
T=   [1  1 -1 1  1 -1   1 -1 -1 1 -1  -1 -1  1  1  1 1 -1
      1 -1  1 1 -1  1  -1  1 -1 1 -1   1 -1 -1  1 -1 1 -1];
% New NN. Transfer function can be tansig or purelin, but
% it would not help.
net=newff([-1 1;-1 1;0 1;0 1;0 1],[2],{'hardlims'});
net.trainFcn='traingdx';
net.trainParam.epochs=2000;
% Start learning
fprintf('===>Start NN training.\n');
net=train(net,P,T);
% Get NN output
Y=sim(net,P);
% See if it is OK or not.
disp('==>See result from NN learning');
T-Y
if sum(sum(T-Y))==0
    disp('Good job, no GA search required.')
else    
  disp('Type <CR> to continue');
  pause

% Use of GA to find those weights and bias of NN
  MUL_factor=1; MIN_offset=1000;
  Emin=MIN_offset; 
  LO=-500*ones(1,12); HI=500*ones(1,12);
  range=[LO;HI];
  obj_fcn='auto_func';
  bit_n = 100;		% Bit number for each input variable
  gen_no = 80;	% Number of generations
  popuSize = 56;		% Population size
  xover_rate = 0.97;	% Crossover rate (BETTER)
  mutate_rate = 0.1;	% Mutation rate (BETTER)
  elite=1;IC=[];
  % call GA 
  tic
  [popu, popu_real, fcn_value, upper, average, lower, ...
    BEST_popu, popuSize, gen_no, para] = GA_genetic(obj_fcn,...
    range, IC, elite, gen_no, popuSize);
  tc=toc/60;
  fprintf('==> Computation time is (%.2f) minutes.\n',tc); 
  % Get weights and bias
  net.IW{1,1}=[popu_real(1,1:5);popu_real(1,6:10)];
  net.b{1}=[popu_real(1,11);popu_real(1,12)];
  Y=sim(net,P);
  disp('==>See result from GA search');
  T-Y
  if sum(sum(T-Y))~=0
  % NN training again
    net.trainParam.epochs=1000;
    net=train(net,P,T);
    Y=sim(net,P);
    disp('==>See result from 2nd-time NN learning');
    T-Y
  end;
  if sum(sum(T-Y))==0
      disp('==>You got a perfect solution, the W and b are:');
      net.IW{1,1}
      net.b{1}
  else
      disp('<<<<< Sorry, try again');
  end;    
end;