Can't use K-Means with some data trees

Dear experts,

I´m trying to cluster some relatively complex data tree using Lunchbox´s K-Means component (28.0 KB). However, it complains with the following error when I directly input such data tree:

1. Solution exception:The points matrix should be rectangular. The vector at position {} has a different length than previous ones. Parameter name: x

Since my data tree may have different number of branches and elements depending on the case, I decided to use a simple Anemone loop to obtain the desired “Good Result”.

Three questions:

  1. Is there any way to directly run K-Means with my complex data tree? Perhaps calling K-Means using some fancy C# code?

  2. When the Anemone loop is integrated into a much bigger GH sketch to perform some Galapagos or TT Toolbox´s brute force optimisation it systematically fails most of the iterations.

  3. Whereas the results from the “Standard Anemone Loop” (blue group) are always the same for different seeds, the results from the “Fast Anemone Loop” (red group) always differ unless a Random Seed of 9 is used instead.

Any help is welcome! Thanks!!

Hi, I’ve found a fancy way to circumvent this K-Means and Anemone problems. Just use the super fast K-Means implementation in OWL (by Mateusz Zwierzycki).

Here it is the updated GH sketch: (39.1 KB)

Hope this helps somebody in the future :slight_smile: