torlarse

I support this. My use case is a point cloud of 3 x 3 arrays, each representing a 9-vector of grayscale pixels. I would like to see e.g the the...

I have started working on this issue.

I personally would support having both versions on pip. Particularly if it is possible to throw an exception if and when the user tries to install more than one version...

For the record, I tried manipulating the ``` if precomputed: min_cluster_samples = 2 ``` condition in `kmapper.py`. To my understanding this defines the minimal number of samples in each cube...

Yes I issued the pull request you are mentioning. I seem to recall that I got a `scikit-learn` chrash if `affinity='precomputed'` and `n_clusters` was not `None`. Does that make sense?...

I certainly can if I am able to get the logic :) 1. By using `affinity=precomputed`, doesn't Scikit require `distance_treshold` to be set? 2. How "safe" is it to manipulate...

ah sorry for not specifying, I have used `sklearn.cluster.AgglomerativeClustering` [1] with single linkage. I am trying to reproduce the work of Carlsson and Gabrielsson [2]. [1]: https://scikit-learn.org/stable/modules/generated/sklearn.cluster.AgglomerativeClustering.html#sklearn.cluster.AgglomerativeClustering [2]: https://arxiv.org/abs/1810.03234

Here is my full clustering section: ``` from sklearn.cluster import AgglomerativeClustering as cl point_cloud_column_variance = np.var(point_cloud, axis=0) point_cloud_normalized = point_cloud / point_cloud_column_variance condensed_distance_matrix = spatial.distance.pdist(point_cloud_normalized) redundant_distance_matrix = spatial.distance.squareform(condensed_distance_matrix) model_vne =...

> It is important to specify here, in case you think otherwise -- hypercubes are always of a fixed interval for a given mapping run. If a given hypercube doesn't...

Thanks, I'll try and implement your suggestion 👍