link_strategy explanation

[Ciro1990]

Hi there,

I have tried to find an explanation for the different link_strategy options, however I have managed to find only this:

link_strategy : {'recursive', 'nonrecursive', 'hybrid', 'numba', 'drop', 'auto'}
    algorithm used to resolve subnetworks of nearby particles
    'auto' uses hybrid (numba+recursive) if available
    'drop' causes particles in subnetworks to go unlinked

Could you help me find the right documentation for such feature or explain me what effect has each option?

One more question. I have found on Wikipedia the theory behind K-tree and B-tree. However, it is not clear when it is more convenient to adopt the first or the latter method. Could you clarify this, please?

Thanks a lot for the support! Ciro

[nkeim]

Look at the "Advanced Linking" and "Performance" tutorials to understand what subnetworks are and how we try to speed them up. All of the link_strategy options except 'drop' implement the same algorithm and are tested to have the same results; they are just implemented/optimized differently. Avoid drop unless you want trackpy to throw out subnetworks instead of trying to solve them.

With few exceptions, you should use a K-tree even if you suspect a B-tree may be more appropriate for your data. The reason is that scipy's K-tree implementation is highly optimized (and written in C). Of course you can always try each and see which one is faster for you.

[Ciro1990]

Thanks a lot for the explanation!