OUT OF MEMORY IN MATLAB

hi all dear i want to run my program that it has very large matrix and when it runs .i see out of memory error.how can i fix that?

 if true
    clc;
    close all;
    clear all;
    t=0:0.0001:900;
    number=200;
%==========================================================================
    synchronization_time=400;
    dysynchronization_time=550;
    c0_value=1.5;
    k=0;
    delay=2.5;
    a=1;
 %-------------------------------------------------------------------------
    zbar=zeros(number,length(t));
    xbar=zeros(number,length(t));
    ybar=zeros(number,length(t));
    x=zeros(number,length(t));
    y=zeros(number,length(t));
    z=zeros(number,length(t));
    xs=zeros(1,length(t));
    ys=zeros(1,length(t));
    s=zeros(1,length(t));
    andaze_s=zeros(1,length(t));
    delta=zeros(1,number);

    zbars=zeros(1,length(t));
    zbars_conj=zeros(1,length(t));
    zbars_tavan2=zeros(1,length(t));
    sum1=zeros(number,length(t));
    sum3=zeros(number,length(t));
    sum4=zeros(number,length(t));

    desync=0;
    delta=zeros(1,number);

    init=2*rand(1,2*number)-1;
    f=.05;
%======================================================

    c0=0;

    for l=1:number
     z(l,1)=init(1,l)+j*init(1,2*l);
    end
    xs(1)=1;
    ys(1)=1;

    omega=normrnd(((2*pi)/5),0.1,number,1);
%===================================================

    for l=1:length(t)-1
      if (l>(synchronization_time*10000)&&l<((synchronization_time+20)*10000))
        c0=c0_value; 
      end
      if l>(dysynchronization_time*10000)
        k=100;
     end
%===================================================
     sum2=0;
     if l>(26000)
       for d=1:number  
         sum2=(x(d,l)+j*y(d,l))+sum2;
       end
       zbars(l)=sum2/number;
       zbars_conj(l)=conj(zbars(l-(delay*10000)));
       zbars_tavan2(l)=(zbars(l)^2);

       s(l)=k*zbars_tavan2(l)*zbars_conj(l);
       xs(l)=real(s(l));
       ys(l)=imag(s(l));
       andaze_s(l)=abs(s(l));
     end

     for m=1:number
      delta(m)=1/(1+(f^2)*((abs((number/2)-m))^2));
      z(m,l+1)=z(m,l)+.0001*((a+j*omega(m)-(abs((z(m,l)))^2))*(z(m,l))+c0*zbar(m,l)+(delta(m)*(s(l))));
      x(m,l+1)=real(z(m,l+1));
      y(m,l+1)=imag(z(m,l+1));
     end

    for h=1:number
     for g=1:number
        sum1(h,l)=(1/(1+f*abs(h-g)))*((z(g,l)))+sum1(h,l);
     end
     zbar(h,l+1)=sum1(h,l)/number;
    end

    if l==5525
     hg=0
    end

%==========================================================================

end
figure
color_fig=rand(1,120);
for f=1:30
 plot(x(f,:),'color',[color_fig(f) color_fig(2*f) color_fig(3*f)]);
 hold on;
end
figure
plot(andaze_s);
figure
plot(real(s));
figure
plot(imag(s));
end

[Jan]

Please explain which line causes the error, the versions of Matlab and the operating system (64 bit?) and the amount of installed RAM.

Please omit "clear all", because it wastes time for no benefit.

Answer 1

[Andreas Goser]

I was tempted to suggest more RAM, but seriously in most applications I see, users simply do not plan / think of "How much data do I need to answer my question?"

There are different application with different patterns.

1. Measured data to be read and postprocessed. Do I really need 1 ms sampled data to be able to identify frequencies of 50 Hz?

2. Simulations: Do I really need all the results of all channels of the simulation? Or isn't it enough to see the last 10 seconds of this one signal?

...

You get the picture...

Answer 2

[Image Analyst]

Heeding Andreas's advice, why don't you try your code with this:

t=0:1:900; % Perhaps this resolution is good enough.

instead of

t=0:0.0001:900; % Way more resolution than is needed.

Answer 3

[Jan]

Each of the 9 matrices need 14'400'000'000 Bytes. So about 14*9=126 Gigabyte free memory are the absolute minimum, 172GB would be more reliable.

The usual suggestion is:

• Install more RAM
• Care for a proper pre-allocation
• Install more RAM
• Run a 64-bit version of Matlab
• Install more RAM
• Close other processes
• Install more RAM
• Use the smallest data type, which represents your data exactly, e.g. a single occupies 4 bytes instead of 8 for a double.
• Install more RAM

[Walter Roberson]

Jan forgot to mention that you should consider installing more RAM!

On the other hand, if you need 128 gigabyte, you probably need to re-examine your algorithm.

[Jan]

Oh, sorry, you are right, Walter, I forgot this important detail. And besides the constructive advice to increase the RAM, reducing the resolution of the examination is an obvious strategy also, therefore I've voted for Andreas' and IA's suggestions.