Remove For Loops to Optimize Performance

cvx_begin

y = 0;

n = 1000;

variables x;

minimize y;

parfor i = 1:n

y = -sum(log(1-x^2)) - sum(log(4 + rand(i)*x));

end

cvx_end

Need to remove for loops and write directly as vectors to speed up optimization algorithm.

[Torsten]

rand(i) generates an (ixi) matrix of uniformly distributed random numbers on [0 1].

I doubt this is what you want.

Answer 1

[Taru]

Hi,

There seems to be a misunderstanding of the rand function as rand(i) gives a i*i square matrix of uniformly distributed random numbers between [0 1]. I understand that you might want it to generate n random numbers. So, you can just use rand(1,n) and vectorise it as:

y = -sum(log(1-x^2)) - sum(log(4 + rand(1,n)*x));

Without using any loop of sorts.