Question about the weighting factor of Exponential Weighted Moving Average (EWMA/EMA)

Asked Jun 19 '19 at 08:06

Active Jun 19 '19 at 14:26

Viewed 246 times

Hiii, I have one question about the weighting factor of EMA.

As I learned, Exponential Weighted Moving Average has a weighting factor, Lamda, and its formula is:

S(t) = Lamda * Y(t) + (1-Lamda) * S(t-1)

Is there any methods to choose the lamda adaptively? Not to set the fixed la, choose the 'best lamda' adaptively. I want to apply it to some online learning problems, and it's better to choose the best weighting factor automatically.

Is there any related paper or github code?

Thank you soooooo much!^^

edited Jun 19 '19 at 14:26

Richard Hardy

67,272

asked Jun 19 '19 at 08:06

user145177

There surely is. Google "exponential smoothing in R" or Python or your favourite software package. – Richard Hardy Jun 19 '19 at 14:26
Hi Richard Hardy, thank you for your response. But I want to find a hyperparameter-free exponential smoothing method. Do you know any paper about it? Thank you^^ – user145177 Jun 20 '19 at 02:21
I think most of the exponential smoothing functions will estimate the $\lambda$ for you. Otherwise it is extremely easy to write your own code for an exponential smoother for a given $\lambda$; estimation of $\lambda$ is the hard part. Try out function ets from the forecast package in R. It estimates $\lambda$ and even facilitates (optionally) automated selection between different versions of exponential smoothing models. – Richard Hardy Jun 20 '19 at 07:46
Related: Hyperparameter-free method for Moving Average/ Exponential smoothing?. – Richard Hardy Jun 20 '19 at 09:17
Ok, thank you so much. I will try it now. – user145177 Jun 20 '19 at 10:58

Question about the weighting factor of Exponential Weighted Moving Average (EWMA/EMA)

0 Answers0

Linked