Choose function too slow

Dear All,

i try to run

index=nchoosek(1:200, 10);

it talks a lot of time

is there exist any code that can be more faster.

Thanks in Advance.

Answer 1

[Image Analyst]

Look at the number of rows it would return:

>> factorial(200)./(factorial(200-10)* factorial(10))
ans =
   NaN
>> factorial(200)
ans =
   Inf

What do you really want to do. I doubt you need to do it this way. Perhaps rand or randperm might work for you.

[Walter Roberson]

The number of possibilities is about 2.2E+16

Answer 2

[Walter Roberson]

An array that large cannot be done on any platform supported by MATLAB. It is 22451004309013280 outputs, each 10 entries long. Now, 200 is a number that fits within 8 bits so in theory the individual entries could be a single uint8 (a byte each); that still requires 2^(57.64) bytes of memory. No released x86 CPU has been designed that uses more than 48 bits of memory addressing, so your problem is 2^(ceil(57.64)-48) = 1024 times larger than is possible on any supported platform at this time.

Answer 3

[Jan]

FEX: VChooseK is much faster than Matlab's nchoosek. But for your job a really large memory is required. So think twice if you really need such a giantic array.

Answer 4

[Hassan]

Thanks for your replay.

when i use VChooseK, Error in compiler for matlab R2012a

No supported SDK or compiler was found on this computer. For a list of supported compilers, see http://www.mathworks.com/support/compilers/R2012a/win64.html

Error using mex (line 206) Unable to complete successfully.

[Walter Roberson]

So see that URL and load in a compiler and SDK.

[Jan]

You find several threads about installing a compiler in this forum.

Did you follow my advice to think twice?