description is quite vague, so not every case is obvious
for example, is it possible that a dash may be the same length as triple space? is situation when only dashes(3) and longspaces(7) are included? the proportion is quite close. Is it given that theres such float number unit, so every sequence up to 2unit is 1 unit long, up to 5unit - 3 units long, others are 7 units long?

I am i a great pain as i want to resolve this on Go and there's no clustering in standart lib(
The lack of explicit prior distributions seems a problem for me: for example, if theres much more dots then dashes, then one may expect distribution like this:
...|
..|||..
.|||||...|
|||||||.|||(as sum of normals),
thus border is weighted mean, but k-means predicts border as unweighted mean

same note about a priori dot is in previous kata, most likely it is needed to solve 111000111, and doesn help to resolve the issue

somehow i have avg 85158.47 instead of 85158.49 at 40((
Maybe someone can help?
PS i found out myself that changing float32 to float64 helped