[selmanozleyen]

fixes: #1244

will add more description but I want to start somewhere @sarajimenez, @FrancescaDr these are all the things I could find in https://github.com/Jieran-S/hitchhikers_guide_to_spatial_project/blob/spatial_clustering/notebooks/Spatially_aware_clustering/sac.ipynb.

Anything else needed? If everything is there, how does this state look like for you?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.52%. Comparing base (dc82c73) to head (67bac38).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1276      +/-   ##
==========================================
+ Coverage   78.95%   79.52%   +0.56%     
==========================================
  Files          63       65       +2     
  Lines        9439     9617     +178     
  Branches     1574     1599      +25     
==========================================
+ Hits         7453     7648     +195     
+ Misses       1469     1458      -11     
+ Partials      517      511       -6     

Files with missing lines	Coverage Δ
src/squidpy/gr/_autok.py	100.00% <100.00%> (ø)
src/squidpy/gr/_cluster_auto_k.py	100.00% <100.00%> (ø)
src/squidpy/gr/_clusterers.py	89.74% <100.00%> (+1.86%)	⬆️
src/squidpy/gr/_nhood.py	62.64% <100.00%> (+1.08%)	⬆️
src/squidpy/gr/_niche.py	88.88% <100.00%> (+0.33%)	⬆️
src/squidpy/types.py	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[grst]

I think a Bayesian GMM could also be very interesting for the same purpose: https://scikit-learn.org/stable/modules/generated/sklearn.mixture.BayesianGaussianMixture.html

This class allows to infer an approximate posterior distribution over the parameters of a Gaussian mixture distribution. The effective number of components can be inferred from the data.

The downside is that it's quite slow - not sure how hard it would be to make a fast GPU implementation for that @Intron7.

[grst]

Does this have to be tied to the GaussianMixture? We deliberately split the niche calling procedure such that arbitrary embeddings and clustering algos can be combined... In principle, why not use the same approach for Leiden?

[selmanozleyen]

I didn't want to define a clusterer protocol publicly. And to have a hard-coded list of clusterers I wanted some opinions first. Bayesian GMM is a good hint

[grst]

this is embarrassingly parallel. Probably the clustering algorithm has some sort of threading, too, but did you evaluate how well it scales to many cores? Maybe parallelizing at this level could make sense.

[selmanozleyen]

I think we rely on the clustering algorithm to be utilizing the cores as much as possible so that any other threading would oversubscribe only. Also we can only parallelize runs if tolerance is 0, ie we don't early stop.

But from my small benches I didn't see much improvement than 1.5x tbh. And BLAS was already enough.

[grst]

I was planning to open a separate issue the next days to discuss a public "extensibility" interface for the niche algos. We can discuss it again in that context.

[grst]

The runs only differ at the level of random seeds. Would it make sense to combine this with e.g. bootstrapping?

[selmanozleyen]

I am not sure, it sounds interesting but also sounds out of scope for this PR. wdyt @sarajimenez @timtreis ?

[sarajimenez]

I do think it's out of the scope of this PR, we're interested in the changes between Ks not across runs. Bootstrapping would be interesting in case we want to add significance. Am I reading this correct, @grst ?