Hello Pascal,
Am I understanding you correct? If I have sine waves dataset and the
sine period contains ~2000 data points then I should set swarm over >6000?
Thank you
On 11/23/2015 05:02 PM, Pascal Weinberger wrote:
No. Sorry but the most hated, but best answer here is:
It depends....
On that your data looks like. A good thing I use to do is just plot the entire
series and look at it to try to identify the patterns myself. Then look at how
many data points your identified (lowest frequency) pattern takes, and try to
get some (min. 2/3) of them in your swarm.
So if you see the longest patterns having ~2000 data points, try to swarm over
>6000 :)
That way you'll get the best model Parameters for all the sequences in your
data :)
Best,
Pascal
