0
$\begingroup$

I'm trying to figure out if the noise in my experiments is gaussian or not.

I don't have time series data. I first configure my system and then run some software application's and measure the average performance and power consumption over N iterations. So at the end I don't have time series data. Instead, I have the aggregate results over several iterations.

There is significant variance in the results between iterations of experiments. I want to check whether the noise in the performance and power consumption is gaussian.

I'm confused about how to test this for my data. Furthermore, if the noise is gaussian. What would be the best way to get the de-noised performance and power consumption for the experiments. I would be extremely grateful for any help in this matter.

$\endgroup$
0
$\begingroup$

I would try and histogram the data you have based on a metric you care about. If you measure the average performance and power consumption over N iterations, it means that you have the performance and power consumption N times. When you histogram this (how many iterations had such and such performance and\or power consumption) you may start to get a feel about the distribution of the noise etc. The size if N will be important to asses whether the distribution of your noise is Gaussian in nature (which it is most likely) or different, or mixed. De-noising is a general approach that usually use some information on the measurement at hand, is the "signal" smooth, or sparse in some representation etc. So it dependents on the character of your signal.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.