Error when cutting Paris dendrogram about scikit-network

[tbonald]

Thank you very much for your feedback.
This has been fixed (please check the develop branch; will be available in the next release).

[ninasachdev]

Thank you so much for your prompt response! I updated the functions in the hierarchy module, but am running into another error:

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-50-33674d51a39e> in <module>
----> 1 labels = cut_straight_dev(dendrogram, n_clusters=100, return_dendrogram=True)

<ipython-input-7-adb9d8c1d297> in cut_straight_dev(dendrogram, n_clusters, threshold, sort_clusters, return_dendrogram)
     92             cluster[n + t] = cluster.pop(i) + cluster.pop(j)
     93 
---> 94     return get_labels_dev(dendrogram, cluster, sort_clusters, return_dendrogram)

<ipython-input-7-adb9d8c1d297> in get_labels_dev(dendrogram, cluster, sort_clusters, return_dendrogram)
     19         current_cluster_new = len(clusters)
     20         for i, j, height, _ in dendrogram:
---> 21             i_new = cluster_index.pop(int(i))
     22             j_new = cluster_index.pop(int(j))
     23             if i_new != j_new:

KeyError: 59725

Since there are several inf values in the height column of my dendrogram, reorder_dendrogram is comparing two rows with the same inf height, and thus the order of the dendrogram doesn't change. The cut variable in cut_straight() is also equal to inf.

This KeyError occurs the first time if i_new != j_new: is True in get_labels(). Could this have to do with how my dendrogram is ordered?

Please let me know if I can clarify anything!

[tbonald]

Hard to debug without the dendrogram variable. Is there a way to share it?

[ninasachdev]

Yes I just emailed the dendrogram to you, let me know if you have issues with opening it!

[tbonald]

Hi @ninasachdev
I seems that there is an inversion in your dendrogram (on row 5663, the merge is done at a height lower that those of the clusters), causing the error in the cut.
So the problem is probably in the generation of the dendrogram. Could you please send me the original graph?
Thanks.

[tbonald]

I'm closing this issue. Feel free to reopen it if needed.