# Matlab problem - index out of bounds because numel(temp3)=0

1. Jan 29, 2007

### cammo12

Hi,

I am using a Matlab program called f2matlab, which converts fortran 90 code to matlab code.

But I keep on getting the following error:

Attempted to access temp3(1); index out of bounds because numel(temp3)=0.

Does anyone know what this error relates to? :uhh:

2. Jan 29, 2007

### Dr Transport

Your variable temp3 is an array and as your program loops through the code it is finding that your array index is too large, i.e. you have declared the variable to be an array of length 2 and your code says to loop through it 3 or more times.

3. Jan 29, 2007

### cammo12

Thanks.

It is strange because the program will work fine with some fortran code I have, but then not with other code. I just don't know why! Back to pulling my hair out...

4. Jan 29, 2007

### J77

Specifically, it's trying to access the first element of temp3. However, temp3 is empty - no elements, hence numel(temp3)=0.

Somewhere in your code, you probably want to initialise temp3 - even if it's with zeroes.

5. Jan 29, 2007

### cammo12

It is throwing up the error on this line of code:

temp3=temp3(temp3>temp4); temp2=temp3(1)-1;

what does temp3(temp3>temp4) mean?

Sorry about all these stupid questions

6. Jan 29, 2007

### J77

It means to make a new temp3 which contains elements such that temp3>temp4, where temp3 (the array that will be overwritten by a new temp3) and temp4 are arrays.

7. Jan 29, 2007

### neurocomp2003

always remember tha tmatlab works with vectors/matrices in regards to its variables.
The equation that may be causing the error is

temp2=temp3(1)-1

Separate that line into 2 lines...and see if it is...if it is then print temp3 (using fprintf or disp, both are your friends, fprintf>disp)

Last edited: Jan 29, 2007
8. Jan 29, 2007

### cammo12

neurocomp2003 - yes I have tried this and it is temp2=temp3(1)-1 that is the problem.

print temp3 = 0.

still trying to figure out why temp3 is zero.... back to the code :grumpy:

9. Jan 29, 2007

### J77

Is that zero or empty?

If zero, there's no problem calling temp3(1).

If empty, there is a problem.

ps: try help keyboard for quick and convenient debugging needs.

10. Jan 29, 2007

### cammo12

temp3 = Empty matrix: 1-by-0

The code I am using has been tried and test by thousands of people - you can download from the matlab website.

But for some reason I am getting this error with some of the code I am using.

11. Jan 29, 2007

### J77

What's the code?

12. Jan 29, 2007

### cammo12

13. Jan 29, 2007

### J77

OK - glad it's fixed - I also see from the comments that the f2matlab function isn't very robust.

14. Jan 26, 2010

### Christiane

Hi,

I'm from brazil.
I'm using a Matlab program called MMFM, which calculates the number of states in a Markov chain fluid.

But I keep on getting the following error:

??? Attempted to access flambda(4); index out of bounds because numel(flambda)=3.

Error in ==> MMFM>birthdeath at 89
lambda_i=flambda(i);

Error in ==> MMFM at 48
[tjump, state] = birthdeath(npoints, fl, fm, A);

My code is going in loop, how can I fix this?

Thanks.

The code:
function [state,R]=MMFM(pkts,a,N,M,npoints)

%Parâmetros:
%pkts=Série de tráfego a ser modelada
%a=proveniente da modelagem da função de autocovariância
%M=A cadeia de markov possui M+1 estados. Variando de zero a M.
%m=média da série
%v=variância da série
%Inputs:
%pkts,a,N,M,npoints
%Outputs:
%state, R

m=mean(pkts);
v=var(pkts);

%Modela as probabilidades do tráfego através de uma distribuição binomial
beta=a/(1+(N*((N*m)^2))/(M*N*v));
alfa=a-beta;
A=(N*v)/(N*m)+(N*m)/M;
%Inicializa a matriz de transições de estados
R=zeros(M+1);
%loop para cadeia de markov birth-death
M=M+1; %O programa considera estados {1,2,...,M+1}
for i=1:M
if i<M
R(i,i+1)=(M-i)*alfa;
end
if i>1
R(i,i-1)=(i-1)*beta;
end
R(i,i)=0;
for j=1:M
if abs(i-j)>1
R(i,j)=0;
end
end
end
%Nascimentos
nasc=M*alfa;
fl=[nasc:-alfa:alfa];
%Mortes
morte=M*beta;
fm=[beta:beta:morte];
[tjump, state] = birthdeath(npoints, fl, fm, A);
eixo=[1:npoints];
plot(eixo,state); %%%plota mudanças de estados sem considerar tempos de
%transições
figure;

[i,j]=hist(state,120);
y=i/npoints;
plot(j,y);
hold on
[p,k]=hist(pkts,120);
yy=p/npoints;
plot(k,yy,'r-');
media=mean(pkts);
tamanho=length(pkts);
poiss=poissrnd(media,tamanho,1);
[pp,kk]=hist(poiss,120);
yy=pp/npoints;
plot(kk,yy,'y-');
legend('MMFM','Real','Poisson');
figure;
plot(tjump,state);
xlabel('Tempo(Segundos)');

%Chama programa birthdeath
function [tjump, state] = birthdeath(npoints, flambda, fmu,A)
% BIRTHDEATH generate a trajectory of a birth-death process
%
%[tjump, state] = birthdeath(npoints,flambda,fmu)
%Inputs: npoints - length of the trajectory
%Outputs: tjump - jump times
%state - states of the embedded Markov chain

i=1; %initial value, start on level i
tjump(1)=0; %start at time 0
state(1)=i; %at time 0: level i
for k=2:npoints
% compute the intensities
lambda_i=flambda(i);
mu_i=fmu(i);

time=-log(rand)./(lambda_i+mu_i); % Inter-step times:
% Exp(lambda_i+mu_i)-distributed

if rand<=lambda_i./(lambda_i+mu_i)
i=i+1; % birth
else
i=i-1; % death
end %if
state(k)=i;
tjump(k)=time;
end %for i

tjump=cumsum(tjump); %cumulative jump times

state=state-1;
state=state*A;

15. Jan 26, 2010