How can i put the conditional statement ( If-Then-else) on the specific elements (Rows/columns) of a matrix in Matlab Using Neural Network Tool?

How can i put the conditional statement ( If-Then-else) on the specific elements (Rows/columns) of a matrix in Matlab Using Neural Network Tool? Basically I am working on classification using neural Network tool in matlab. My out put only shows in matrix like : output =

    1.2723
    0.0123
    0.0022
But I want the output in statement like "Class1" Or "Class2"
i.e. If the matrix is like this then output is this....

For example: I have two different matrices like: output1 =

    1.2723
    0.0123
    0.0022

output2 =

    0.2460
   -0.0521
    1.1062

%I want to set the ranges of the elements on behalf of some conditions like:

 If The first element of the  first row of matrix Output1 is >= 0.6 and <=1.5

AND

If The 2nd element of the 2nd row of matrix Output1 is >= -0.49 and <=0.49

AND

If The 3rd element of the 3rd row of matrix Output1 is >= -0.49 and <=0.49

THEN

Display = Class1

%for Matrix Output2:

I want to set the ranges of the elements on behalf of some conditions like:

If The 1st element of the 1st row of matrix Output2 is >= -0.49 and <=0.49

AND

If The 2nd element of the 2nd row of matrix Output2 is >= -0.49 and <=0.49

AND

If The 3rd element of the 3rd row of matrix Output2 is >= 0.6 and <=1.5

THEN

Display = Class2

Please anyone can send me the code/explanation. how can I set these condition using If-Then-else or any other statement?

Regards:

Safdar Hayat

[phone]

Skype:

jugnu2008.safdar

Iqra University Islamabad, Pakistan

Answer 1

[Greg Heath]

If you have 4 classes, your target matrix columns should be columns of eye(4). The relationship between class indices 1:4 and the target matrix is

 trueclassindices = vec2ind(target)
 target           = ind2vec(trueclassindices)

Example:

 >> trueclassindices = [ 1 1 2 4 3 4  ]
 >> target           = full( ind2vec(trueclassindices))

 target =   1     1     0     0     0     0     
            0     0     1     0     0     0  
            0     0     0     0     1     0 
            0     0     0     1     0     1

If

 output =[ 0.5142 0.49114  0.18198      0     0.25288    0.13838
          0.03653 0.1455   0.4168   0.14434   0.18459    0.066341
          0.31642 0.22294  0.26658  0.38809   0.54807    0.35987
          0.13285 0.14041  0.13464  0.46757   0.014469   0.43541]

then

 assignedclasses =   1     1     2     4     3     4

Hope this helps.

Thank you for formally accepting my answer

PS. The unit sum output is a consistent estimate of the input-conditional posterior class probabilities (See any pattern-recognition text).

Greg

[Safdar Hayat Khaksar]

Thanks for your Answer Greg But still I have facing problem.

like basically I have three classes 1,2 & 3:

I have many target sets as:

t1  =    [0;0;0];

  
t2  =    [0;0;0];

  
t3  =    [0;0;0];

  
t4  =    [0;0;0];

  
t5  =    [0;0;0];  

t6  =    [0;0;0];  

t7  =    [0;0;0];  

.

.

.

.  

t342  =    [0;0;0];

%These all belongs to class1.

And

t343  =    [1;0;0];  

t344  =    [1;0;0];  

.

.

.

.  

t485  =    [1;0;0];

%These all belongs to class2.

And

t486  =    [0;0;1];  

t487  =    [0;0;1];  

.

.

.  

t560  =    [0;0;1];

%These all belongs to class3.

Now when I train my network by giving total 560 examples and 560

targets then my network generate the output according to the

target set like:

      output =

         -0.0045

          0.0056

          0.0314

      %Belongs to class1

      output =

          0.9779

         -0.0082

          0.0425

      %Belongs to class2

      output =

          0.0078

          0.0099

          1.0794

      %Belongs to class3

Now I want to display a message "class1" instead of matrix output .

for example if the output is resembled with

output =

     -0.0045

      0.0056

      0.0314 

Then Display “Class1”

Means If The value of 1st element of the 1st row of matrix "Output" is in rage of >= -0.49 and <=0.49

AND

If The value of 2nd element of the 2nd row of matrix "Output" is in rage of >= -0.49 and <=0.49

AND

If The  value of 3rd element of the 3rd row of matrix "Output" is in rage of >= 0.6 and <=1.5

THEN

Display “ Class”

And want to repeat the same conditions for all outputs of all classes.

Basically I want to put the condition (range values) on the value of each element of output matrix and then display a “message” on behalf of those conditions.

Hope you will understand my question in detail.

Thank you for your moral support.

Regards:

Safdar Hayat

[Greg Heath]

I understand your question.

My answer is to change your target to have {0,1} unit vectors, modify your code and train a new classifier.

Hope this helps.

Greg

[Safdar Hayat Khaksar]

Hi: Dear Greg Health !

Thanks for your moral support but your answer did not help me.

I have solved my problem by self hit and try.... :-p

I have used just targets matrix indices and then applied the loop simply as follow:

for output=sim(net, test)

    X=output;

if  ((X(1,1)>= -0.499) && (X(1,1)<= 0.5))&& ((X(2,1)>= -0.499) && (X(2,1)<= 0.5)) && ((X(3,1)>= -0.499) && (X(3,1)<= 0.5))

          display('Class 1');

elseif  ((X(1,1)>= 0.5) && (X(1,1)<= 1.5)) && ((X(2,1)>= -0.499) && (X(2,1)<= 0.5)) && ((X(3,1)>= -0.499) && (X(3,1)<= 0.5))

      display('Class 2');

elseif ((X(1,1)>= -0.499) && (X(1,1)<= 0.5)) && ((X(2,1)>= -0.499) && (X(2,1)<= 0.5)) &&  ((X(3,1)>= 0.5) && (X(3,1)<= 1.5))

      display('Class 3');

else

      display('no math found');

end

end

% X(1,1) represent first element of first row and first column of matrix X.

%X(2,1) represent first element of 2nd row and first column of matrix X.

%X(3,1) represent first element of 3rd row and first column of matrix X.

• _Yahoooo... !!!

Thanks alot_*