search

EthoML / VAME

Variational Animal Motion Embedding - A tool for time series embedding and clustering
https://ethoml.github.io/VAME/
GNU General Public License v3.0
2 stars 0 forks source link

Community detection not working #16

Closed vinicvaz closed 3 weeks ago

vinicvaz commented 1 month ago

I tried to run the demo.py pipeline with vame.community(config, show_umap=False, cut_tree=2) post processing and I got the following error:

line 448, in get_cohort_community_labels
    clust = np.asarray(communities_all[i])
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (4,) + inhomogeneous part.

Looks like the communities_all array is not homogeneous for this case... Looking to this variable I have the following inhomogeneous array: [[[3, 10, 7, 5, 13], [6, 1, 8, 2], [9, 4, 14], [12, 11, 0]]]

Steps to reproduce the above behavior:

  1. Run the demo.py from tests branch uncommenting the vame.community(config, show_umap=False, cut_tree=2) line.
  2. Set the config.yaml project file to use max_epochs=20

What may be causing this issue? Can it be related to the amount of epochs I'm using to test?

vinicvaz commented 1 month ago

@luiztauffer @bendichter @katiekly fyi

bendichter commented 1 month ago

Can we get away with keeping it as a list of lists?