a=rand(400), this function generates 400*400 random matrix a,how could generate matrix a with specified mean and variance.

a=rand(400), this function generates 400*400 random matrix a,how could generate matrix a with specified mean and variance in matlab.

[Image Analyst]

Is this homework?

[vipul utsav]

what do you mean? and what you feel

[Image Analyst]

If a poster honestly admits it's homework, then we will give hints rather than full solutions so they can't be accused of plagiarizing, and won't feel guilty about turning in someone else's code as their own. It appears that you imply this is not your homework, so don't worry about it.

Answer 1

[Andrei Bobrov]

m - your mean

s - your variance

size_matrix - size of your array < a x b > , size_matrix = [a,b];

fun1 = @(m,s,size_matrix) m + sqrt(s)*randn(size_matrix)

use function fun1, eg:

>> fun1(1,2,[3,2])

please read about randn

if you have Statistics Toolbox, use function normrnd:

>>   normrnd(m,sqrt(s),size_matrix);

[vipul utsav]

fun1 = 20 + sqrt(0.6)*randn(400); here i got mean=19.9995 it is ok but i don'i get variance ok. i got it 0.0018

[vipul utsav]

fun1 = 20 + sqrt(0.6)*randn(400) why this function does not give right answer?

[vipul utsav]

 fun1= normrnd(20,sqrt(0.6),[400 400]);
fun1 = 20 + sqrt(0.6)*randn(400);
both function gives proper mean but it does not give proper variance 
i.e  nearest 0.6

[Image Analyst]

Why use randn() or normrnd()? vipul, in your original post you hinted at a uniform distribution because you used rand(). The suggestion to look at normal distributions is good though. But which distribution do you want? It seems strange that you would switch from one to another since they are so different. Have you given the shape of the distribution any thought?

[vipul utsav]

this is different question compared to you answered (homogeneous area...)

and i got solution

[Image Analyst]

Yes I know I answered a different question from you in a different discussion but the comment I made here was related to this discussion. Usually people care about what shape their noise distribution takes and I was just wondering why you don't, because it has a huge impact on the results. But whatever, as long as you think you have a solution....