moadbahron commented 2 years ago

Hi ,

The # of realizations when its 1000 , the code take a lot of time to get results, while the Matlab CPU is ~50% and RAM ~1.6 GB which is normal , but when I change the #of setups and #of realizations 20,100 respectively , the results are shown , but the curves are not smooth as your paper results . untitled1

Is my laptop is not suitable for your code or what it seems the issue?

emilbjornson commented 2 years ago

Hi!

You need a sufficient number of random data points on the curves to get smooth results from a Monte Carlo simulation. The number of points equals the number of scenarios. We used 200 points to get relatively smooth curves. 20 points as in your case is not enough.

I think you can get away with 100 realizations per scenario but you need 200 scenarios. The run time should be proportional to the number of scenarios. So 200 takes 10 times more time than 20.

It can take a few days to generate highly accurate simulation results.

17 aug. 2022 kl. 19:36 skrev moadbahron @.***>:

Hi ,

The # of realizations when its 1000 , the code take a lot of time to get results, while the Matlab CPU is ~50% and RAM ~1.6 GB which is normal , but when I change the #of setups and #of realizations 20,100 respectively , the results are shown , but the curves are not smooth as your paper results .

Is my laptop is not suitable for your code or what it seems the issue?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.

moadbahron commented 2 years ago

Hi,

"Number of setups with random UE locations" in other words Is it the # of trails of random positions of UE location . and 200 is the min or optimum value for the curves to be smooth .

emilbjornson commented 2 years ago

Yes, we concluded that 200 setups was sufficient to get visually smooth curves, but more setups will always lead to even smoother results so it is an accuracy vs. complexity tradeoff.

Even if the curves look smooth, there will be an error margin that one must take into consideration when comparing closely spaced curves. We are using the Monte Carlo method to create an empirical CDF and there are ways to characterize the estimation errors if you want to look deeper into it: https://stats.stackexchange.com/questions/342315/accuracy-of-empirical-cumulative-distribution-function

On 17 Aug 2022, at 20:14, moadbahron @.***> wrote:

Hi,

"Number of setups with random UE locations" in other words Is it the # of trails of random positions of UE location . and 200 is the min or optimum value for the curves to be smooth .

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.

moadbahron commented 2 years ago

Hi again ,

Right now I have running other disktops to run each Fig at your recommendation , But in the meantime I have selected