Penuntun Praktikum JJaarriinnggaann SSyyaarraaff TTiirruuaannfisika.unud.ac.id/diktat/praktikum/praktikum jst.pdfModul Praktikum Jaringan Syaraf Tiruan JST Hebb 1.1 Program Aplikasi

Penuntun Praktikum

JJaarriinnggaann SSyyaarraaff TTiirruuaann DDeennggaann MMaattllaabb GGUUII

IGA Widagda

Fisika FMIPA UNUD

2014

© igawidagda -lab.fisika komputasi unud

1

Modul Praktikum Jaringan Syaraf Tiruan

JST Hebb

1.1 Program Aplikasi Komputer JST Hebb

1.1.1.Rancangan Graphical User Interface (GUI)

1.1.2 Daftar Komponen

No Komponen Properti

1 Panel title Training

2 Axes Tag axes1

3 Axes Tag axes2

4 Axes Tag axes6

5 Axes Tag axes7

6 Axes Tag axes8

7 Push button String open image 1

Tag image1_pushbutton









11

Gambar 1.1 Rancangan GUI program aplikasi JST Hebb

11


2


1.1.3. Kode Program (source code)

1.1.3.1 Function image1_pushbutton _Callback

function image1_pushbutton_Callback(hObject, eventdata, handles) [filename,pathname]=uigetfile('*.jpg;*.bmp;*.png','pilih gambar anda'); handles.image_file = strcat(pathname,filename); %"handles.image_file" var image_file %di set agar bisa diakses dari function %yang lain if exist(handles.image_file) % check if image exists handles.im_original1=imread(handles.image_file); % read in the image axes(handles.axes1); % will display on axes1 imshow(handles.im_original1); else % if image doesn_t exist imshow('file_not_found.jpg') % Display the _Error image_ end guidata(hObject,handles);




12 Static text String Bobot (w) dan Bias (b) hasil training

13 List box String kosong

Tag Show_listbox

14 Push button String start training

Tag start_training

15 Panel Title Testing

16 Axes Tag axes3

17 Axes Tag axes4

18 Push button String open testing image

Tag image_testing_pushbutton

19 Push button String Start testing

Tag start_testing_pushbutton

20 Static text String Hasil Testing


3








1.1.3.4 Function start_training_pushbutton _Callback

function start_training_pushbutton_Callback(hObject, eventdata, handles) s1=rgb2gray(handles.im_original1); s2=rgb2gray(handles.im_original2); s3=rgb2gray(handles.im_original3); s4=rgb2gray(handles.im_original4); s5=rgb2gray(handles.im_original5); %ubah data menjadi bipolar[1,-1] jml_baris=size(s1,1); jml_kolom=size(s1,2);


4


for i=1:1:jml_baris for j=1:1:jml_kolom if s1(i,j)>=200 %putih s1_temp(i,j)=-1; else %hitam s1_temp(i,j)=1; end end end for i=1:1:jml_baris for j=1:1:jml_kolom if s2(i,j)>=200 %putih s2_temp(i,j)=-1; else %hitam s2_temp(i,j)=1; end end end for i=1:1:jml_baris for j=1:1:jml_kolom if s3(i,j)>=200 %putih s3_temp(i,j)=-1; else %hitam s3_temp(i,j)=1; end end end for i=1:1:jml_baris for j=1:1:jml_kolom if s4(i,j)>=200 %putih s4_temp(i,j)=-1; else %hitam s4_temp(i,j)=1; end end end for i=1:1:jml_baris for j=1:1:jml_kolom if s5(i,j)>=200 %putih s5_temp(i,j)=-1; else %hitam s5_temp(i,j)=1; end end end %jml_pola= 5; % 5 citra %ubah matrik jadi jadi matrik kovektor k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s(1,k)=s1_temp(i,j); end end t1=1; %target untuk citra 1 simpan sbg 3 kolom %akhir dalam array s t2=1; t3=1;


5


s(1,k+1)=t1; s(1,k+2)=t2; s(1,k+3)=t3; k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s(2,k)=s2_temp(i,j); end end t1=1; t2=1; %target untuk citra 2 t3=-1; s(2,k+1) = t1; s(2,k+2)= t2; s(2,k+3)=t3; k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s(3,k)=s3_temp(i,j); end end t1=1; %target untuk citra 3 t2=-1; t3=1; s(3,k+1) = t1; s(3,k+2) = t2; s(3,k+3)= t3; k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s(4,k)=s4_temp(i,j); end end t1=1; %target untuk citra 4 t2=-1; t3=-1 s(4,k+1) = t1; s(4,k+2) = t2; s(4,k+3)= t3; k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s(5,k)=s5_temp(i,j); end end t1=-1; %target untuk citra 5 t2= 1; t3= 1; s(5,k+1) = t1; s(5,k+2) = t2; s(5,k+3) = t3; %inisialisasi bobot (w) dan bias (b) jml_pola=size(s,1);


6


jml_input = size(s,2)-3; %-3 krn output/target ada 3 jml_output=3; %inisialisasi botot w dan bias b for j=1:1:jml_output for i=1:1:jml_input w(i,j)=0; end end for j=1:1:jml_output b(j)=0; end for n=1:1:jml_pola for i=1:1:jml_input x(i) = s(n,i); %%x(i) = s(i) end for j=1:1:jml_output t(n,j)=s(n,jml_input+j); %target t % t(n,2)=s(n,jml_input+2); %target t y(n,j)=t(n,j); % y1 % y(n,2)=t(n,2); % y2 end %perbaiki bobot dan bias for j=1:1:jml_output for i=1:1:jml_input w(i,j) = w(i,j) + x(i)*y(n,j); % w_baru =w_lama+x(i)* y handles.w_final(i,j)=w(i,j); end b(j)=b(j)+y(n,j); %b_baru= b_lama + y handles.b_final(j)=b(j); end end %%tampilkan bobot dan bias hasil pelatihan di command window for i=1:1:jml_input for j=1:1:jml_output disp(strcat('w',num2str(i),num2str(j),' = ',num2str(w(i,j)))); end end for j=1:1:jml_output disp(strcat('b',num2str(j),' = ',num2str(b(j)))); end %Tampilkan bobot dalam list box n=1; for i=1:1:jml_input %index dari array bobot w for j=1:1:jml_output k(n)=str2num(strcat(num2str(i),num2str(j))); w_vek(n)=w(i,j); %ubah w jadi vektor w_vek n=n+1; end end %tampilkan bobot bj, sbg elemen terakhir listbox set(handles.show_listbox,'String',strcat('w',num2str(k'),' = ',num2str(w_vek'))); for j=1:1:jml_output bobot=strcat('b',num2str(j),' = ',num2str(b(j))); list_box_old=cellstr(get(handles.show_listbox,'String')); list_box_new= [list_box_old;{bobot}]; set(handles.show_listbox,'String',list_box_new) ;


7


end guidata(hObject,handles);

1.1.3.5 Function image_testing_pushbutton _Callback

function image_testing_pushbutton_Callback(hObject, eventdata, handles) [filename,pathname]=uigetfile('*.jpg;*.bmp;*.png','Select an Image File'); handles.image_file = strcat(pathname,filename); %"handles.image_file" var image_file %di set agar bisa diakses dari function %yang lain if exist(handles.image_file) % check if image exists handles.im_testing=imread(handles.image_file); % read in the image axes(handles.axes3); % will display on axes1 imshow(handles.im_testing); else % if image doesn_t exist imshow('file_not_found.jpg') % Display the _Error image_ end guidata(hObject,handles);

1.1.3.6 Function start_testing_pushbutton _Callback

function start_testing_pushbutton_Callback(hObject, eventdata, handles) s=rgb2gray(handles.im_testing); %ubah data menjadi bipolar[1,-1] jml_baris=size(s,1); jml_kolom=size(s,2); for i=1:1:jml_baris for j=1:1:jml_kolom if s(i,j)>=200 %putih s_temp(i,j)=-1; else %hitam s_temp(i,j)=1; end end end %ubah matrik citra 2 dimensi s_temp jadi vektor 1 dimensi s k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s_vektor(k)=s_temp(i,j); end end jml_input = k; jml_output =3; net(1)=0; net(2)=0; net(3)=0; for j=1:1:jml_output for i=1:1:jml_input x(i)=s_vektor(i); net(j)=net(j)+x(i)*handles.w_final(i,j); end net(j)=net(j)+handles.b_final(j); end axes(handles.axes4); if net(1)>=0 && net(2)>=0 && net(3)>=0


8


imshow(handles.im_original1); else if net(1)>=0 && net(2)>=0 && net(3)<0 imshow(handles.im_original2); else if net(1)>=0 && net(2)<0 && net(3) >=0 imshow(handles.im_original3); else if net(1)>=0 && net(2)<0 && net(3) <0 imshow(handles.im_original4); else if net(1)<0 && net(2)>=0 && net(3) >=0 imshow(handles.im_original5); else imshow(handles.im_original5); end end end end end

1.1.4 Hasil Eksekusi (RUN) program

Hasil eksekusi program JST Hebb dapat dilihat dalam Gambar 1.2.

Cara kerja program :

- Buka file citra atau gambar dengan menekan tombol open image 1 sampai open

image 5

- Klik tombol Start Training, setelah menunggu beberapa saat sistem akan

menampilkan bobot (w) dan bias (b) hasil pelatihan list box

- Buka file citra atau gambar dengan menekan tombol open testing image

- Klik tombol Start Testing, selanjutnya akan ditampilkan citra hasil testing yang

cocok dengan citra yang diuji.

Gambar 1.2 Hasil Eksekusi program JST Hebb


9


2.1 Program Aplikasi Komputer JST Hebb (model 2)

2.1.1.Rancangan Graphical User Interface (GUI)

2.1.2 Daftar Komponen


1 Panel Title Training

2 Axes Tag axes1

3 Axes Tag axes2

4 Axes Tag axes3

5 Push button String open


6 Push button String <

Tag before_pushbutton

7 Push button String >

Tag next_pushbutton

8 Push button String Start Training

Tag start_trainning_pushbutton



10 Push button String Start Testing


Gambar 1.3 Rancangan GUI program aplikasi JST Hebb model 2


10


2.1.3. Kode Program (source code)


function image1_pushbutton_Callback(hObject, eventdata, handles) current_dir=pwd; %getting current directory handles.img_dir=uigetdir(current_dir,'Select directory'); %Selecting directory %where image saved %Reading image files at one folder if exist(handles.img_dir) cd(handles.img_dir); %change directory D = dir('*.jpg'); %list all image-file in directory handles.img_original = cell(1,numel(D));%create 1 X numel(D) cell array of empty matrices % numel(D)=number of element in directory D for i = 1:numel(D) [handles.img_original{i},handles.fs{i}] = imread(D(i).name); %save into array 'img_original' end axes(handles.axes1); for i=1:numel(D) imshow(handles.img_original{i}); end handles.number_of_image=numel(D); handles.number_of_max=numel(D); else h=msgbox('File not found', 'Warning','error'); end cd(current_dir);%return into current directory guidata(hObject,handles);

2.1.3.2 Function before_pushbutton _Callback

function before_pushbutton_Callback(hObject, eventdata, handles) axes(handles.axes1); % will display on axes1 i=handles.number_of_image; if i==1 % before_pushbutton.Enable='off'; imshow(handles.img_original{1}); %handles.number_of_img=1; else imshow(handles.img_original{i}); handles.number_of_image = handles.number_of_image-1;


11 Static text String Database Citra

12 Static text2 String Weight (w) dan Bias (b)


Tag show_listbox




11


end guidata(hObject,handles);

2.1.3.3 Function next_pushbutton _Callback

function next_pushbutton_Callback(hObject, eventdata, handles) axes(handles.axes1); % will display on axes1 i=handles.number_of_image; if i == handles.number_of_max imshow(handles.img_original{i}); else imshow(handles.img_original{i}); handles.number_of_image = handles.number_of_image+1; end guidata(hObject,handles);

2.1.3.4 Function start_trainning_pushbutton _Callback

function start_trainning_pushbutton_Callback(hObject, eventdata, handles) num_wav=size(handles.img_original,2); %number of wav files for i=1:1:num_wav s(i,:,:)=rgb2gray(handles.img_original{i}); end jml_citra=size(s,1); jml_baris=size(s,2) jml_kolom=size(s,3) %ubah citra asli bw menjadi bipolar[1,-1] for k=1:1:jml_citra for i=1:1:jml_baris for j=1:1:jml_kolom if s(k,i,j)>200 %putih s_temp(k,i,j)=-1; else %hitam s_temp(k,i,j)=1; end end end end %ubah citra 'source' format matrik baris kolom(i,j) jadi format baris %citra 1 for m=1:1:jml_citra k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s(m,k)=s_temp(m,i,j); end end end jml_input=k %tentukan matrik target handles.jml_target = 3; % 2^handles.jml_target jml_digit=handles.jml_target; m=0;


12


i=2^handles.jml_target-1; for k=1:num_wav x=de2bi(i,jml_digit); %desimal to biner, tampilkan sebanyak 'jumlah digit' handles.target(k,:)=x; i=i-1; end %ubah target menjadi bipolar for k=1:num_wav for i=1:jml_digit if handles.target(k,i)==0 handles.target(k,i)=1; else handles.target(k,i)=-1; end end end %inisialisasi bobot (w) dan bias (b) jml_pola=size(s,1) %jml_input = size(s,2) jml_output=handles.jml_target; %inisialisasi botot w dan bias b for j=1:1:jml_output for i=1:1:jml_input w(i,j)=0; end end for j=1:1:jml_output b(j)=0; end for n=1:1:jml_pola for i=1:1:jml_input x(i) = s(n,i); %%x(i) = s(i) end for j=1:1:jml_output %t(n,j)=s(n,jml_input+j); %target t % t(n,2)=s(n,jml_input+2); %target t y(n,j)=handles.target(n,j); % y1 % y(n,2)=t(n,2); % y2 end %perbaiki bobot dan bias for j=1:1:jml_output for i=1:1:jml_input w(i,j) = w(i,j) + x(i)*y(n,j); % w_baru =w_lama+x(i)* y handles.w_final(i,j)=w(i,j); end b(j)=b(j)+y(n,j); %b_baru= b_lama + y handles.b_final(j)=b(j); end end %Tampilkan bobot dalam list box n=1; for i=1:1:jml_input %index dari array bobot w for j=1:1:jml_output k(n)=str2num(strcat(num2str(i),num2str(j))); w_vek(n)=w(i,j); %ubah w jadi vektor w_vek n=n+1; end end


13


%tampilkan bobot bj, sbg elemen terakhir listbox set(handles.show_listbox,'String',strcat('w',num2str(k'),' = ',num2str(w_vek'))); for j=1:1:jml_output bobot=strcat('b',num2str(j),' = ',num2str(b(j))); list_box_old=cellstr(get(handles.show_listbox,'String')); list_box_new= [list_box_old;{bobot}]; set(handles.show_listbox,'String',list_box_new) ; end guidata(hObject,handles);

2.1.3.3 Function image_testing_pushbutton _Callback

function image_testing_pushbutton_Callback(hObject, eventdata, handles) [filename,pathname]=uigetfile('*.jpg','Pilih file citra(*.jpg)'); handles.img_file = strcat(pathname,filename); if exist(handles.img_file) % check if image exists axes(handles.axes2); % will display on axes1 handles.img_testing=imread(strcat(pathname,filename)); %read image imshow(handles.img_testing); else % if image doesn_t exist h=msgbox('file not found', 'Error','error'); end handles.pathname=pathname; handles.filename=filename; guidata(hObject,handles);

2.1.3.4 Function start_testing_pushbutton _Callback

function start_testing_pushbutton_Callback(hObject, eventdata, handles) s=rgb2gray(handles.img_testing); %ubah data menjadi bipolar[1,-1] jml_baris=size(s,1); jml_kolom=size(s,2); for i=1:1:jml_baris for j=1:1:jml_kolom if s(i,j)>200 %putih s_temp(i,j)=-1; else %hitam s_temp(i,j)=1; end end end s_temp %ubah matrik citra baris-kolom s_temp jadi matrik baris k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s_vektor(k)=s_temp(i,j); end end jml_input = k jml_output=handles.jml_target for i=1:jml_output net(i)=0; end size(handles.w_final) for j=1:1:jml_output


14


for i=1:1:jml_input x(i)=s_vektor(i); net(j)=net(j)+x(i)*handles.w_final(i,j); end net(j)=net(j)+handles.b_final(j) end for i=1:jml_output if net(i)>=0 net(i)=1 else net(i)=-1 end end jml_pola=size(handles.target,1); cocok=0; axes(handles.axes3); for i=1:jml_pola if handles.target(i,:)==net() imshow(handles.img_original{i}); cocok=1; handles.wav_hasil=strcat('wav',num2str(i),'.wav'); break; end end if cocok==0 h=msgbox('Tidak ada yang cocok','warning'); end guidata(hObject,handles);

2.1.4 Hasil Eksekusi (RUN) program

Hasil eksekusi program JST Hebb dapat dilihat dalam Gambar 1.4.


- Buka file citra atau gambar dengan menekan tombol open pada panel training

Gambar 1.4 Hasil Eksekusi program JST Hebb model 2


15


- Klik tombol Start Training, setelah menunggu beberapa saat sistem akan

menampilkan bobot (w) dan bias (b) hasil pelatihan pada list box

- Buka file citra atau gambar dengan menekan tombol open pada panel testing



3. Latihan

1. Buatlah percobaan dengan menggunakan program aplikasi JST Hebb model 2

dengan 5 buah citra pola huruf untuk proses pelatihan. Selanjutnya ujilah atau

testing sistem dengan citra huruf yang cacat atau mengandung noise.

2. Buatlah percobaan dengan memakai program aplikasi JST Hebb model 2 dengan

5 buah citra wajah untuk proses pelatihan atau training. Kemudian testing system

dengan citra wajah cacat atau mengandung noise


1


JST Perceptron

2.1.Rancangan Graphical User Interface (GUI)

11 21

Gambar 2.1 Rancangan GUI program aplikasi JST Perceptron


2


2.2 Daftar Komponen



2 Axes Tag axes1

3 Axes Tag axes2

4 Axes Tag axes3






Tag next_pushbutton




10 Static text1 String Alpha

11 Static text2 String Tetha

12 Edit text1 String kosong

Tag alpha_edit

13 Edit text2 String kosong

Tag tetha_edit

14 Push button String Open testing image





3


2.3. Kode Program (source code)

2.3.1 Function image1_pushbutton _Callback

current_dir=pwd; %getting current directory handles.img_dir=uigetdir(current_dir,'Select directory'); %Selecting directory %where image saved %Reading wav files at one folder if exist(handles.img_dir) cd(handles.img_dir); %change directory D = dir('*.jpg'); %list all wav-file in directory handles.img_original = cell(1,numel(D));%create 1 X numel(D) cell array of empty matrices % numel(D)=number of element in directory D for i = 1:numel(D) [handles.img_original{i},handles.fs{i}] = imread(D(i).name); %save into array 'img_original' end axes(handles.axes1); for i=1:numel(D) imshow(handles.img_original{i}); end handles.number_of_image=numel(D); handles.number_of_max=numel(D); else h=msgbox('File not found', 'Warning','error'); end cd(current_dir);%return into current directory guidata(hObject,handles);

2.3.2 Function before_pushbutton _Callback

axes(handles.axes1); % will display on axes1 i=handles.number_of_image; if i==1 % before_pushbutton.Enable='off'; imshow(handles.img_original{1}); %handles.number_of_img=1; else imshow(handles.img_original{i});




Tag show_listbox


25 Static text4 String Hasil Testing


4


handles.number_of_image = handles.number_of_image-1; end guidata(hObject,handles);

2.3.3 Function next_pushbutton _Callback

function next_pushbutton_Callback(hObject, eventdata, handles) axes(handles.axes1); % will display on axes1 i=handles.number_of_image; if i == handles.number_of_max imshow(handles.img_original{i}); else imshow(handles.img_original{i}); handles.number_of_image = handles.number_of_image+1; end guidata(hObject,handles);

2.3.4 Function start_trainning_pushbutton _Callback

num_file=size(handles.img_original,2); %number of files for i=1:1:num_file s(i,:,:)=rgb2gray(handles.img_original{i}); end jml_citra=size(s,1); jml_baris=size(s,2) jml_kolom=size(s,3) %ubah citra asli bw menjadi bipolar[1,-1] for k=1:1:jml_citra for i=1:1:jml_baris for j=1:1:jml_kolom if s(k,i,j)>200 %putih s_temp(k,i,j)=-1; else %hitam s_temp(k,i,j)=1; end end end end %ubah citra 'source' format matrik baris kolom(i,j) jadi format baris(vektor) for m=1:1:jml_citra k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s(m,k)=s_temp(m,i,j); end end end jml_input=k %tentukan matrik target handles.jml_target = 3; % 2^handles.jml_target jml_digit=handles.jml_target; m=0; i=2^handles.jml_target-1; for k=1:num_file x=de2bi(i,jml_digit); %desimal to biner, tampilkan sebanyak 'jumlah digit' handles.target(k,:)=x;


5


i=i-1; end %ubah target menjadi bipolar for k=1:num_file for i=1:jml_digit if handles.target(k,i)==0 handles.target(k,i)=1; else handles.target(k,i)=-1; end end end alpha=str2num(get(handles.alpha_edit,'String')); %learning rate tetha=str2num(get(handles.tetha_edit,'String')); %threshold %inisialisasi bobot (w) dan bias (b) jml_pola=size(s,1) %jml_input = size(s,2) jml_output=handles.jml_target; %inisialisasi botot w dan bias b for j=1:1:jml_output for i=1:1:jml_input w(i,j)=0; end end for j=1:1:jml_output b(j)=0; end berhenti = false; jml_iterasi = 0; while berhenti==false jml_iterasi=jml_iterasi+1; berhenti=true; for n=1:1:jml_pola for i=1:1:jml_input x(i) = s(n,i); %x(i) = s(i) end %t(n)=s(n,jml_input+1); %target %y(n,j)=handles.target(n,j); % y1 for j=1:1:jml_output %t(n,j)=s(n,jml_input+j); %target t % t(n,2)=s(n,jml_input+2); %target t t(n,j)=handles.target(n,j); % y1 % y(n,2)=t(n,2); % y2 end % sigma=0; for j=1:1:jml_output sigma=0; for i=1:1:jml_input sigma=sigma+x(i)*w(i,j); end y_in(j)=b(j)+sigma; if y_in(j)>tetha y(j)=1; end


6


if y_in(j)>=-tetha && y_in(j)<=tetha y(j)=0; end if y_in(j)<-tetha y(j)=-1; end %jika output(y) tidak sesuai target(t) maka perbaiki %bobot w dan bias b if y(j) ~= t(n,j) for j=1:1:jml_output for i=1:1:jml_input w(i,j)=w(i,j)+alpha*t(n,j)*x(i); handles.w_final(i,j)=w(i,j); end b(j)=b(j)+alpha*t(n,j); handles.b_final(j)=b(j); berhenti=false; end end end end end %%tampilkan bobot dan bias hasil pelatihan dalam command window editor for j=1:1:jml_output for i=1:1:jml_input disp(strcat('w',num2str(i),num2str(j),' = ',num2str(w(i,j)))); end disp(strcat('b = ',num2str(b(j)))); end for n=1:1:jml_input %index dari array bobot w i(n)=n; end %Tampilkan bobot dalam list box n=1; for i=1:1:jml_input %index dari array bobot w for j=1:1:jml_output k(n)=str2num(strcat(num2str(i),num2str(j))); w_vek(n)=w(i,j); %ubah w jadi vektor w_vek n=n+1; end end set(handles.show_listbox,'String',strcat('w',num2str(k'),' = ',num2str(w_vek'))); %tampilkan bobot bj, sbg elemen terakhir listbox for j=1:1:jml_output bobot=strcat('b',num2str(j),' = ',num2str(b(j))); list_box_old=cellstr(get(handles.show_listbox,'String')); list_box_new= [list_box_old;{bobot}]; set(handles.show_listbox,'String',list_box_new) ; end j_iterasi=strcat('jml iterasi = ',num2str(jml_iterasi)); %tampilkan jumlah iterasi sbg elemen terakhir listbox list_box_old=cellstr(get(handles.show_listbox,'String')); list_box_new= [list_box_old;{j_iterasi}]; set(handles.show_listbox,'String',list_box_new) ; guidata(hObject,handles);


7


2.3.5 Function image_testing_pushbutton _Callback

function image_testing_pushbutton_Callback(hObject, eventdata, handles) [filename,pathname]=uigetfile('*.jpg;*.bmp;*.png','Select an Image File'); handles.image_file = strcat(pathname,filename); %"handles.image_file" var image_file %di set agar bisa diakses dari function %yang lain if exist(handles.image_file) % check if image exists handles.im_testing=imread(handles.image_file); % read in the image axes(handles.axes2); % will display on axes1 imshow(handles.im_testing); else % if image doesn�t exist imshow('file_not_found.jpg') % Display the �Error image� end guidata(hObject,handles);

2.3.6 Function start_testing_pushbutton _Callback

function start_testing_pushbutton_Callback(hObject, eventdata, handles) s=rgb2gray(handles.im_testing) %ubah data menjadi bipolar[1,-1] jml_baris=size(s,1); jml_kolom=size(s,2); for i=1:1:jml_baris for j=1:1:jml_kolom if s(i,j)>=200 %putih s_temp(i,j)=-1; else %hitam s_temp(i,j)=1; end end end %ubah matrik citra 2 dimensi s_temp jadi vektor 1 dimensi s k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s_vektor(k)=s_temp(i,j); end end jml_input = k jml_output=handles.jml_target for i=1:jml_output net(i)=0; end size(handles.w_final) for j=1:1:jml_output for i=1:1:jml_input x(i)=s_vektor(i); net(j)=net(j)+x(i)*handles.w_final(i,j); end net(j)=net(j)+handles.b_final(j) end for i=1:jml_output if net(i)>=0 net(i)=1 else


8


net(i)=-1 end end jml_pola=size(handles.target,1); cocok=0; axes(handles.axes3); for i=1:jml_pola if handles.target(i,:)==net() imshow(handles.img_original{i}); cocok=1; handles.wav_hasil=strcat('wav',num2str(i),'.wav'); break; end end if cocok==0 h=msgbox('Tidak ada yang cocok','warning'); end guidata(hObject,handles);

2.4 Hasil Eksekusi (RUN) program

Hasil eksekusi program JST Perceptron dapat dilihat dalam Gambar 2.2.


- Buka file citra atau gambar dengan menekan tombol open pada panel training

- Masukkan nilai alpha (misal : 0.7) dan nilai tetha (misal : 0.3). Klik tombol

Start Training, setelah menunggu beberapa saat sistem akan menampilkan

bobot (w) dan bias (b) hasil pelatihan pada list box

- Buka file citra atau gambar dengan menekan tombol open pada panel testing



Gambar 2.2 Hasil Eksekusi program JST Perceptron


9


2.5 Latihan

1. Buatlah percobaan dengan menggunakan program aplikasi JST Perceptron

dengan 5 buah citra pola huruf untuk proses pelatihan. Selanjutnya ujilah atau

testing sistem dengan citra huruf yang cacat atau mengandung noise.

2. Buatlah percobaan dengan memakai program aplikasi JST Perceptron dengan 5

buah citra sidik jari untuk proses pelatihan atau training. Kemudian testing

system dengan citra wajah cacat atau mengandung noise


1


JST Heteroassociative Memory (HM)


3.2 Daftar Komponen

11 31

Gambar 3.1 Rancangan GUI program aplikasi JST HM


2



2.3.1 Function image1_pushbutton _Callback

current_dir=pwd; %getting current directory handles.image_dir=uigetdir(current_dir,'Select directory'); %Selecting directory %where image saved



2 Axes Tag axes1

3 Axes Tag axes2






Tag next_pushbutton


Tag before2_pushbutton


Tag next2_pushbutton




11 Push button String Open testing image







Tag show_listbox


16 Static text4 String Hasil Testing


3


%Reading multiple image files at one folder if exist(handles.image_dir) cd(handles.image_dir); %change directory D = dir('*.jpg'); handles.im_original = cell(1,numel(D));% number element for i = 1:numel(D) handles.im_original{i} = imread(D(i).name); %save image data into array 'im_original' end axes(handles.axes1); %imshow(imcell{1}); for i=1:numel(D) imshow(handles.im_original{i}); end handles.number_of_image=numel(D); handles.number_of_max=numel(D); else h=msgbox('file not found', 'Warning','error'); end cd(current_dir);%return into current directory guidata(hObject,handles);

2.3.2 Function before_pushbutton _Callback

i=handles.number_of_image; if i==1 imshow(handles.im_original{1}); %handles.number_of_img=1; else imshow(handles.im_original{i}); handles.number_of_image = handles.number_of_image-1; end guidata(hObject,handles);

2.3.3 Function next_pushbutton _Callback

function next_pushbutton_Callback(hObject, eventdata, handles) i=handles.number_of_image; if i == handles.number_of_max imshow(handles.im_original{i}); %handles.number_of_img=1; else imshow(handles.im_original{i}); handles.number_of_image = handles.number_of_image+1; end guidata(hObject,handles);

2.3.1 Function image1_target_pushbutton _Callback

current_dir=pwd; %getting current directory handles.image_dir=uigetdir(current_dir,'Select directory'); %Selecting directory %where image saved %Reading multiple image files at one folder if exist(handles.image_dir) cd(handles.image_dir); %change directory D = dir('*.jpg'); handles.im_original = cell(1,numel(D));% number element for i = 1:numel(D) handles.im_original{i} = imread(D(i).name); %save image data into array 'im_original'


4


end axes(handles.axes2); %imshow(imcell{1}); for i=1:numel(D) imshow(handles.im_original{i}); end handles.number_of_image=numel(D); handles.number_of_max=numel(D); else h=msgbox('file not found', 'Warning','error'); end cd(current_dir);%return into current directory

guidata(hObject,handles);

2.3.4 Function before2_pushbutton _Callback

i=handles.number_of_image_target; if i==1 imshow(handles.im_original_target{1}); %handles.number_of_img=1; else imshow(handles.im_original_target{i}); handles.number_of_image_target = handles.number_of_image_target-1; end guidata(hObject,handles);

2.3.5 Function next2_pushbutton _Callback

function next_pushbutton_Callback(hObject, eventdata, handles) i=handles.number_of_image_target; if i == handles.number_of_max_target imshow(handles.im_original_target{i}); %handles.number_of_img=1; else imshow(handles.im_original_target{i}); handles.number_of_image_target = handles.number_of_image_target+1; end guidata(hObject,handles);

2.3.4 Function start_trainning_pushbutton _Callback

num_im=size(handles.im_original,2); %number of images for i=1:1:num_im s(i,:,:)=im2bw(handles.im_original{i}); end %s(1,:,:) %size(s) for i=1:1:num_im s_target(i,:,:)=im2bw(handles.im_original_target{i}); end %s_target(1,:,:) %size(s_target) jml_citra=size(s,1); jml_baris=size(s,2); jml_kolom=size(s,3);


5


%ubah citra asli bw menjadi bipolar[1,-1] for k=1:1:jml_citra for i=1:1:jml_baris for j=1:1:jml_kolom if s(k,i,j)>0 %putih s_temp(k,i,j)=-1; else %hitam s_temp(k,i,j)=1; end end end end %ubah citra "target" bw menjadi bipolar[1,-1] handles.jml_citra_target=size(s_target,1); handles.jml_baris_target=size(s_target,2); handles.jml_kolom_target=size(s_target,3); for k=1:1:handles.jml_citra_target for i=1:1:handles.jml_baris_target for j=1:1:handles.jml_kolom_target if s_target(k,i,j)> 0 %putih s_target_temp(k,i,j)=-1; else %hitam s_target_temp(k,i,j)=1; end handles.s_target_bipolar(k,i,j)=s_target_temp(k,i,j); end end end %handles.s1_target_bipolar %ubah citra 'source' format matrik baris kolom(i,j) jadi format baris %citra 1 for m=1:1:jml_citra k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; source(m,k)=s_temp(m,i,j); end end end %ubah citra 'target' format matrik baris kolom(i,j) jadi format baris %citra target 1 for k=1:1:handles.jml_citra_target m=0; for i=1:1:handles.jml_baris_target for j=1:1:handles.jml_kolom_target m=m+1; target(k,m)=s_target_temp(k,i,j); end end end %hitung bobot w dengan perkalian (inner product) matrik s dan t


6


jml_pola=size(source,1); jml_input = size(source,2); jml_output=size(target,2); %handles.jml_baris_target*handles.jml_kolom_target; handles.w=0; for i=1:1:jml_pola handles.w= handles.w+source(i,:)'*target(i,:); end %%tampilkan bobot dan bias hasil pelatihan dalam command window %for j=1:1:jml_output % for i=1:1:jml_input % disp(strcat('w',num2str(i),num2str(j),' = ',num2str(handles.w(i,j)))); % end %end %Tampilkan bobot dalam list box %n=0; %for j=1:1:jml_output %index dari bobot w % for i=1:1:jml_input % n=n+1; % in(n)=i; % jn(n)=j; % wn(n)=handles.w(i,j); % end %end %set(handles.show_listbox,'String',strcat('w',num2str(in'),num2str(jn'),' = ',num2str(wn'))); guidata(hObject,handles);

2.3.5 Function image_testing_pushbutton _Callback

function image_testing_pushbutton_Callback(hObject, eventdata, handles) [filename,pathname]=uigetfile('*.jpg;*.bmp;*.png','Select an Image File'); handles.image_file = strcat(pathname,filename); %"handles.image_file" var image_file %di set agar bisa diakses dari function %yang lain if exist(handles.image_file) % check if image exists handles.im_testing=imread(handles.image_file); % read in the image axes(handles.axes3); % will display on axes1 imshow(handles.im_testing); else % if image doesn_t exist imshow('file_not_found.jpg') % Display the _Error image_ end guidata(hObject,handles);


function start_testing_pushbutton_Callback(hObject, eventdata, handles) s=im2bw(handles.im_testing); %ubah data menjadi bipolar[1,-1] jml_baris=size(s,1); jml_kolom=size(s,2); for i=1:1:jml_baris for j=1:1:jml_kolom if s(i,j)>0 %putih s_temp(i,j)=-1; else %hitam s_temp(i,j)=1; end


7


end end s_temp %ubah matrik citra baris-kolom s_temp jadi matrik baris k=0; for i=1:1:jml_baris for j=1:1:jml_kolom k=k+1; s_vektor(k)=s_temp(i,j); end end jml_input = k; jml_output=handles.jml_baris_target*handles.jml_kolom_target; y_temp=s_vektor*handles.w; for j=1:1:jml_output if y_temp(j)>0 y(j)=1; else if y_temp(j)<=0 y(j)=-1; else if y_temp(j)==0 y(j)=0; end end end end %bandingkan output dengan target axes(handles.axes4); for k=1:handles.jml_citra_target if y==handles.s_target_bipolar(k,:); imshow(handles.im_original{k}); break; end end


Hasil eksekusi program JST Perceptron dapat dilihat dalam Gambar 2.2.


8



- Buka file citra atau gambar pelatihan (source image(s)) dengan menekan

tombol open pada panel training

- Buka file citra atau gambar (target image(t)) dengan menekan tombol open

pada panel training

- Lakukan proses training dengan menekan tombol Start Training

- Buka file citra atau gambar testing dengan menekan tombol open pada panel

testing



2.5 Latihan

1. Buatlah percobaan dengan menggunakan program aplikasi JST HM dengan 5

buah citra wajah untuk proses pelatihan. Selanjutnya ujilah atau testing sistem

dengan citra huruf yang cacat atau mengandung noise.

2. Buatlah percobaan dengan memakai program aplikasi JST HM dengan 5 buah

citra sidik jari untuk proses pelatihan atau training. Kemudian testing sistem

dengan citra sidik jari cacat atau mengandung noise

Gambar 3.2 Hasil Eksekusi program JST HM


1


JST Back Propagation (BP)


4.2 Daftar Komponen

11 41

Gambar 4.1 Rancangan GUI program aplikasi JST BP


2





2 listbox String

Tag data_trainning_listbox

3 Push button String open data trainning

Tag data_trainning_pushbutton

4 listbox String

Tag data_wb_listbox

5 Push button String open weight - bias

Tag data_wb_pushbutton

6 Static text String Jumlah hidden unit

7 Edit text String

tag jml_hidden_edit

7 Static text String Learning Rate

8 Edit text String

tag learning_rate_edit

9 Static text String weight(w)-bias(b) after trainning

10 listbox String

Tag wb_result_listbox


Tag start_training_pushbutton


13 listbox String

Tag data_testing_listbox

14 Push button String open data testing

Tag data_testing_pushbutton

15 listbox String

Tag result_listbox




16 Static text String Physics, Unud, Bali


3


4.3.1 Function data_trainning_pushbutton _Callback

[filename,handles.pathname]=uigetfile('*.xls;*.xlsx','Select data trainning'); handles.data_trainning_file = strcat(handles.pathname,filename); if exist(handles.data_trainning_file) % check if files exists handles.data_trainning=xlsread(handles.data_trainning_file); % read data trainning in 'excels' file end set(handles.data_trainning_listbox,'String',num2str(handles.data_trainning)); guidata(hObject,handles);

4.3.2 Function data_wb_pushbutton _Callback

[filename,pathname]=uigetfile('*.xls;*.xlsx','Select data trainning'); handles.wb_file = strcat(pathname,filename); if exist(handles.wb_file) % check if files exists handles.wb=xlsread(handles.wb_file); % read data trainning in 'excels' file end set(handles.data_wb_listbox,'String',num2str(handles.wb)); guidata(hObject,handles);

4.3.3 Function start_training_pushbutton _Callback

jml_baris=size(handles.data_trainning,1); jml_kolom=size(handles.data_trainning,2); jml_pola=size(handles.data_trainning,1); jml_input=size(handles.data_trainning,2)-1; jml_output=size(handles.data_trainning,2)-jml_input; jml_hid=str2num(get(handles.jml_hidden_edit,'String')); alpa=str2num(get(handles.learning_rate_edit,'String')); %reading data source for n=1:1:jml_pola for i=1:1:jml_input s(n,i)= handles.data_trainning(n,i); end end s %reading data target for n=1:1:jml_pola for i=1:1:jml_output t(n,i)=handles.data_trainning(n,jml_input+i); end end t jml_input jml_hid size(handles.wb) %handles.wb(2,4) %reading data weight and bias %reading data weight hidden layer for i=1:1:jml_input for j=1:1:jml_hid v(i,j)=handles.wb(i,j); end end v


4


%reading data bias hidden layer for i=1:1:jml_hid v0(i)=handles.wb(jml_input+1,i); end v0 %reading data weight output layer for i=1:1:jml_hid for j=1:1:jml_output w(i,j)=handles.wb(jml_input+1+i,j); end end w %reading data bias output layer for i=1:1:jml_output w0(i)=handles.wb(jml_input+1+jml_hid+1,i); end w0 eps=0.05; tot_err=0; jml_iterasi=0; stop=false; %********Mulai Pelatihan********** while stop==false %%tot_err=0; for n=1:1:jml_pola for j=1:1:jml_hid tot_j=0; for i=1:1:jml_input x(i)=s(n,i); tot_j=tot_j+x(i)*v(i,j); end z_in(j)=v0(j)+tot_j; %%z[j]=1/(1+exp(-(z_in[j]))); %FUNGSI BINER SIGMOID z(j)=(2/(1+exp(-(z_in(j)))))-1; %FUNGSI BIPOLAR SIGMOID end for k=1:1:jml_output tot_k=0; for j=1:1:jml_hid tot_k=tot_k+z(j)*w(j,k); end y_in(k)=w0(k)+tot_k; %y[k]=1/(1+exp(-y_in[k])); %FUNGSI BINER SIGMOID y(k)=(2/(1+exp(-y_in(k))))-1; %FUNGSI BIPOLAR SIGMOID end %BACKPROPAGATION %tot_err=0; for k=1:1:jml_output %fx=1/(1+exp(-y_in[k])); %biner sigmoid %diff=fx*(1-fx);//turunan pertama fx=(2/(1+exp(-y_in(k))))-1; %Bipolar sigmoid diff=0.5*(1+fx)*(1-fx); %turunan pertama t1(k)=t(n,k); %target tot_err=tot_err+(t1(k)-y(k))^2; %total error kuadrat del_k(k)=(t1(k)-y(k))*diff; %error %tot_err=tot_err+pow(del_k[k],2);


5


for j=1:1:jml_hid delta_w(j,k)=alpa*del_k(k)*z(j); %delta bobot pada unit output end delta_w0(k)=alpa*del_k(k); %delta bias pd unit output end for j=1:1:jml_hid tot_e_k=0; for k=1:1:jml_output tot_e_k=tot_e_k+del_k(k)*w(j,k); end del_in(j)=tot_e_k; %fx=1/(1+exp(-z_in(j))); %biner sigmoid %diff=fx*(1-fx); fx=(2/(1+exp(-z_in(j))))-1; diff=0.5*(1+fx)*(1-fx); del_j(j)=del_in(j)*diff; for i=1:1:jml_input x(i)=s(n,i); delta_v(i,j)=alpa*del_j(j)*x(i); %delta bobot pd unit hidden end delta_v0(j)=alpa*del_j(j); %delta bias pd unit hidden end %UBAH BOBOT DAN BIAS %Ubah bobot unit output for j=1:1:jml_hid for k=1:1:jml_output w(j,k)=w(j,k)+delta_w(j,k); end end %ubah bias unit output for k=1:1:jml_output w0(k)=w0(k)+delta_w0(k); end %ubah bobot unit hidden for i=1:1:jml_input for j=1:1:jml_hid v(i,j)=v(i,j)+delta_v(i,j); end end %ubah bias unit hidden for j=1:1:jml_hid v0(j)=v0(j)+delta_v0(j); end end %for n if tot_err>eps %continue tot_err=0; jml_iterasi=jml_iterasi+1 stop=false; else %stop stop=true; end end %while stop %saving trainning result into excels file v(jml_input+1,:)=v0; %bias unit hiden


6


for i=1:1:jml_hid v(jml_input+1+i,1)=w(i); %bobot unit output end for i=1:1:jml_output v(jml_input+1+jml_hid+i,1)=w0(i); %bias unit output end %simpan bobot w dan bias b dalam excel file if exist(strcat(handles.pathname,'bobot_hasil_backprop.xls')); delete (strcat(handles.pathname,'bobot_hasil_backprop.xls')); xlswrite(strcat(handles.pathname,'bobot_hasil_backprop.xls'),v); else xlswrite(strcat(handles.pathname,'bobot_hasil_backprop.xls'),v); end set(handles.wb_result_listbox,'String',num2str(v)); %%tampilkan bobot dan bias hasil pelatihan dalam comman window %for j=1:1:jml_output % for i=1:1:jml_input % disp(strcat('w',num2str(i),num2str(j),' = ',num2str(handles.w(i,j)))); % end %end %Tampilkan bobot dalam list box %n=0; %for j=1:1:jml_output %index dari bobot w % for i=1:1:jml_input % n=n+1; % in(n)=i; % jn(n)=j; % wn(n)=handles.w(i,j); % end %end %set(handles.wb_result_listbox,'String',strcat('w',num2str(in'),num2str(jn'),' = ',num2str(wn'))); guidata(hObject,handles);

4.3.4 Function data_testing_pushbutton _Callback

[filename,pathname]=uigetfile('*.xls;*.xlsx','Select data trainning');

handles.data_testing_file = strcat(pathname,filename);

if exist(handles.data_testing_file) % check if files exists

handles.data_testing=xlsread(handles.data_testing_file); % read data trainning in

'excels' file

end

set(handles.data_testing_listbox,'String',num2str(handles.data_testing'));

guidata(hObject,handles);


jml_pola=size(handles.data_trainning,1);


7


jml_input=size(handles.data_trainning,2)-1; jml_output=size(handles.data_trainning,2)-jml_input; jml_hid=str2num(get(handles.jml_hidden_edit,'String')); alpa=str2num(get(handles.learning_rate_edit,'String')); %reading data source for i=1:1:jml_input x(i)= handles.data_testing(1,i); end %read weight and bias of trainning wb_testing=xlsread(strcat(handles.pathname,'bobot_hasil_backprop.xls')); %reading data weight of hidden layer , v for i=1:1:jml_input for j=1:1:jml_hid v(i,j)=wb_testing(i,j); end end v %reading data bias hidden layer, v0 for i=1:1:jml_hid v0(i)=wb_testing(jml_input+1,i); end v0 %reading data weight output layer, w for i=1:1:jml_hid for j=1:1:jml_output w(i,j)=wb_testing(jml_input+1+i,j); end end w %reading data bias output layer , w0 for i=1:1:jml_output w0(i)=wb_testing(jml_input+1+jml_hid+1,i); end w0 %proses Testing for j=1:1:jml_hid tot=0; for i=1:1:jml_input tot=tot+x(i)*v(i,j); end z_in(j)=v0(j)+tot; %z(j)=1/(1+exp(-z_in(j))); %biner sigmoid z(j)=(2/(1+exp(-z_in(j))))-1; %bipolar sigmoid end for k=1:1:jml_output tot=0; for j=1:1:jml_hid tot=tot+z(j)*w(j,k); end y_in(k)=w0(k)+tot; %y[k]=1/(1+exp(-y_in[k])); y(k)=(2/(1+exp(-y_in(k))))-1; end set(handles.result_listbox,'String',num2str(y')); %set(handles.wb_result_listbox,'String',list_box_new) ; y


8



Hasil eksekusi program JST BP dapat dilihat dalam Gambar 4.2.


- Buka file data pelatihan (dalam format file excel) dengan menekan tombol

open data trainning pada panel Trainning

- Buka file bobot (w) dan bias (b) (dalam format excel) dengan menekan tombol

open weight-bias pada panel Trainning

- Masukkan Jumlah hidden unit, misal : 4

- Masukkan nilai Learning Rate, misal : 0.2

- Lakukan proses training dengan menekan tombol Start Training

- Buka file testing (format excel) dengan menekan tombol open data testing

pada panel Testing

- Klik tombol Start Testing, selanjutnya akan ditampilkan data hasil testing yang

cocok dengan data pelatihan, yaitu : -0.90633 dibulatkan jadi -1. Jadi cocok

dengan data pada pola pelatihan pertama yaitu :

Data input x1 x2 = -1 -1 maka outputnya -1

Gambar 4.2 Hasil Eksekusi program JST BP kasus Gerbang logika XOR


9


4.5 Latihan

1. Buatlah percobaan dengan menggunakan program aplikasi JST BP dengan data

pelatihan gerbang logika AND.

2. Buatlah percobaan dengan menggunakan program aplikasi JST BP dengan data

pelatihan gerbang logika OR.

Penuntun Praktikum JJaarriinnggaann SSyyaarraaff TTiirruuaannfisika.unud.ac.id/diktat/praktikum/praktikum jst.pdfModul Praktikum Jaringan Syaraf Tiruan JST Hebb 1.1 Program Aplikasi

Documents