Vame.community_analysis() issues

[DmitryKishkinev]

Dear all

I went through all VAME steps up to the community analysis vame.community_analysis() which does not work for me unfortunately.

Perhaps the issue is that we have unequal representation of motif across our videos, for ex, for video-1 we may have motifs [1,2,4] out of say 5 motifs in total used but for video-2 we may have only motifs [1,2,3,5] etc, and my hunch is the current community function does not like it.

Anyway, I believe some people here including some developers here mentioned that the communit_analysis function is not too user friendly atm and might be prone to issues / needs to be improved in the future. In the meantime I believe I can use some custom-written code to post process VAME outputs and do transition matrix and a binary tree plotting etc similar to Luxem et al. 2022 Comm Biol paper. I read the part of Luxem et al. 2022 describing how to build a binary tree using the cost function.

Could anyone point me in the right direction in terms of custom-made community analysis using perhaps own R/Python code (perhaps even sharing some bits of such code) before this issue will be fixed in the vame library. Basically, I have VAME output .npy file (also converted to txt) and a series of motif labels (one column with motif label number on each line representing a frame) and I need code to build transition matrix and then a binary tree as in Luxem et al. 2022.

Any help / pointing in the right direction would be greatly appreciated. My email d (dot) kishkinev (at) keele (dot) ac (dot) uk [pls edit]

Regards
Dmitry