How to avoid the ??? Index exceeds matrix dimensions.

Hai everybody. I m a new student in Matlab and I am doing a small project in Matlab regarding forecasting Temperature inside , Humidity Inside, CFM of ceiling fan in a specific example of indoor house, and Temperature Out and Humidity Out. Now the problem is i couldnt get the excepted FCM to show the memory. Here I am using coding with Differential Hebbian Rule to train the network. Unfortunately I cannot get the answer for the corrupted figures. Please guide me if I am wrong.

clc
%clear all

% DATASET STRUCTURE 

Original=[   % this is a raw dataset of 5 variables and 48 samples

% var1: temp in  var2:humidity in  var3:cfm of ceiling fan  var4:temp out   var5:humidity out  ... 
  29  59  1725  35  55            
  29  59  2535  35  55        
  29  58  3600  35  55        
  30  51  1725  36  57      
  30  49  2535  36  57          
  30  49  3600  36  57      
  36  42  1725  34  52      
  36  42  2535  34  52      
  34  43  3600  34  52      
  36  44  1725  35  54      
  36  44  2535  35  54        
  35  44  3600  35  54        
  35  42  1725  35  57      
  35  44  2535  35  57        
  33  45  3600  35  57              
  32  48  1725  26  71      
  32  48  2535  26  71      
  30  49  3600  26  71      
  30  57  1725  25  78      
  30  57  2535  25  78  
    29  59  3600    25  78
    31  59  1725    22  84
    31  59  2535    22  84
    30  59  3600    22  84
    26  59  1725    29  55
    26  59  2535    29  55
    27  58  3600    29  55
    27  51  1725    29  57
    27  49  2535    29  57
    30  49  3600    29  57
    29  42  1725    30  52
    31  42  2535    30  52
    33  43  3600    30  52
    32  44  1725    30  54
    32  44  2535    31  54
    32  44  3600    31  54
    31  42  1725    30  57
    31  44  2535    30  57
    31  45  3600    29  57
    29  48  1725    28  71
    29  48  2535    28  71
    29  49  3600    28  71
    28  57  1725    27  78
    28  57  2535    27  78
    28  57  3600    27  78
    27  59  1725    27  84
    27  59  2535    27  84
    27  59  3600    27  84];

DataSet=(Original-1)./6;  % we scale down all data within (0,1)

% TRAINING INITIALIZATION

Rec=size(DataSet,1);      % returns Rec=48
Var=size(DataSet,2);      % returns Var=5

alpha=0.02;               % alpha: learning rate
W= repmat(0,Var,Var);     % W is adjacency weights matrix

% TRAINING

start=clock;  % read system time 

for Epoch=1:1000 % for 1000 times we repeat training for all (samples)
 for sample=1:Rec          % loop for learning all samples

       X=DataSet(sample,:);        % pick a sample

       Y=X*W; 

       for j=1:Var            
         for i=1:Var       
            delta=alpha*(Y(j)-X(j))*(Y(i)-X(i));    % differential Hebbian learning rule
            W(i,j)=W(i,j)+delta; 
         end
       end

   end
  end

stop=clock;   % read system time (STOP time)
Run_Time= 60*(stop(5)-start(5))+stop(6)-start(6)       % computational cost 

surf(W)                         % w is the knowledge base (adjacency model)

Memory=DataSet*W;

Error=sqrt(sum(sum((Memory-DataSet).^2))./(Rec*Var))   % all data RMS error

%--------------------------------------------------------------------------

% TESTING

testRecord=  [20  50  1725  35  55];% alages: this is the corrupted sample done by me
testRecord=(testRecord-1)./6;     % standardization by squashing within 0,1 

test=testRecord;

for equ=1:10
    test=test*W;   % for 10 times, we apply the model (W) to the test record to rebuild it
    for fix=1:30   % remeber, every time those certain values within test record must be retained
        test(fix*6-5:fix*6-1)=testRecord(fix*6-5:fix*6-1);
    end
end

test=test*6+1      % back to Original data format

max(abs(Original(1,:)-test))   % now compare the rebuilt record #1 with the original record #1 to get the maximum error

[Image Analyst]

You forgot to give the error message. Copy ALL THE RED TEXT. Don't snip or paraphrase - we need to see the exact error message so we can know what index you're exceeding. Or else find out the error a lot faster than asking us by viewing this link http://blogs.mathworks.com/videos/2012/07/03/debugging-in-matlab/

[Alagesan]

thank you sir. the link really help me out. I will get back without snip or paraphrase.Thank you

[Neural Network]

response to Mr Image Analyst.

If the red text appear like this;

??? Subscript indices must either be real positive integers or logicals.

Error in ==> example at 96
        test(fix*3-5:fix*3-1)=testRecord(fix*3-5:fix*3-1);

when the test record used is

testRecord=  [1000  152  0.35];

testRecord=(testRecord-1)./3;      

test=testRecord;
test(fix*3-5:fix*3-1)=testRecord(fix*3-5:fix*3-1) 

for equ=1:10
    test=test*W;   
    for fix=1:1

what is the error really indicates??? tq

[Walter Roberson]

It is not recommended to use "fix" as a variable name, as "fix" is the name of a MATLAB routine.

You have not indicated what "fix" was initialized to. If it was initialized to 1, then 1*3-5 is 3-5 which is -2 which is not a positive integer and so cannot be a subscript.

[Alagesan]

can anyone tell me the function for fix??

Answer 1

[Image Analyst]

You ask "can anyone tell me the function for fix??" Use k or loopIndex or counter or any other loop index/iteration variable that is not a keyword or reserved function. If in doubt, type "which k" or "whos k" on the command line to see if it's in use already. You might also use parentheses around your index range fix*3-5:fix*3-1 to make it more clear to the reader what you want. For example

5:8-1
5:(8-1)
(5:8)-1

ans =
     5     6     7
ans =
     5     6     7
ans =
     4     5     6     7

[Alagesan]

okie i get it. after i fixed i still cant get their real answer.

here the original data is

original=[27,59,1725,35,55];

so, i need to test the data and get back the real valu in column 5 . Thus I am inserted data

testRecord= [27,59,1725,35, 53]; %look column 5 is corrupted now. Then i used this method...

testRecord=(testRecord-1)./5; % standardization by squashing within 0,1

test=testRecord;

for equ=1:5 test=test*W; for fix=1:1 test(fix*5-4:fix*5-1)=testRecord(fix*5-4:fix*5-1); end end

test=test*5+1; % back to Original data format

max(abs(Original(1,:)-test))

so it must corrected the corrupted column five to give near value. But i couldnt get it... this is my problem...i got another value which is higher then the corrupted.

[Alagesan]

The file that I attached will make you clear. I just do not know how to return back the corrupted value.please guide me guys..

[Image Analyst]

Looks like you posted a separate message for this so we'll continue the discussion there.

[Alagesan]

okie. I separated it. So it wont looks fussy..thank you