The advantage of this smoothing function is that it doesn't need any parameters - it finds the optimal parameters by itself. And still the calculation takes just a second for 100 samples.
This code implements Nadaraya-Watson kernel regression algorithm with Gaussian kernel. The optimal setting of the regression is derived by closed form leave-one-out cross-validation.
Simple to use and fast, thank you.
I just uploaded a partially vectorised version, which should be a bit faster. Now the smoothing takes around 10 seconds to process 2500 samples on my computer. Let me know if you need it to be even faster.
Unfortunately, the calculation for more samples (for example 2500 samples) takes a very long time.
Isn't possible to make it faster?
Improved help text.
The function description was truncated and some tags added.
Screen-shot was added (2).
Removed __MACOSX file from the archive.
Screen-shot was added.