Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: how to vectorize this loop
Replies: 3   Last Post: Apr 19, 2011 9:41 AM

 Messages: [ Previous | Next ]
 Steven Lord Posts: 18,038 Registered: 12/7/04
Re: how to vectorize this loop
Posted: Apr 19, 2011 9:41 AM

"Roger Stafford" <ellieandrogerxyzzy@mindspring.com.invalid> wrote in
message news:ioiqil\$42j\$1@fred.mathworks.com...
> "Mateusz Gos" <webmaster24@wp.pl> wrote in message
> <ioinh4\$f1g\$1@fred.mathworks.com>...

>> hi,
>> i would like to write the following loop in form of a vector, which I am
>> guessing is possible (or not?):
>>
>> n = 1;
>> for n = 1:20
>> R = mean(measurements(1:n));
>> n = n+1;
>> end
>>
>> thanks,
>> mat

> - - - - - - - -
> You should not have the "n = n+1;" line in there. The for-loop with n
> takes care of advancing n. Also the line "n = 1;" is totally unnecessary
> since the for-loop will start at n = 1 no matter how you set n in advance.
>
> As your code stands at present you are taking twenty different mean
> operations, the first one the mean of the single number 1, then the mean
> of the two number 1 and 2, then that of 1, 2, and 3, and so forth. Is
> that what you had in mind? If so, all but the last one of these means
> from 1 to 20 have been overwritten because you placed them in the same
> variable R without indexing it with n.
>
> If all you want is the single mean of the numbers from 1 to 20, then do
> this:
>
> R = mean(1:20);
>
> If you want the twenty different means, do this:
>
> R = zeros(20,1);
> for n = 1:20
> R(n) = mean(1:n);
> end

For this particular problem, since the mean of a vector of values is the sum
of the elements in the vector divided by the number of elements in the
vector, you could use CUMSUM.

% Sample data
measurements = rand(1, 20);
n = 20;

% Engine
R = cumsum(measurements(1:n))./(1:n);

R(5) - mean(measurements(1:5))

--
Steve Lord
slord@mathworks.com