DBSCAN DenseMatrix Requirement

[maximilian-gelbrecht]

Hello,

Right now, if providing DBSCAN with a distance matrix, is has to be a DenseMatrix (or subtype of). I can't see a reason why one needs to be so strict here, it should be fine as long as there is a proper view function defined on the type.

In particular, I tried using it with PairwiseListMatrices but of course there can also be other desirable situation where one does not want to hand over a DenseMatrix to the algorithm.

[alyst]

We have started to relax the input matrix requirements for some algorithms (#129, #138).
It could be, though, that for optimal performance DBSCAN relies on the input matrix being dense.
You are welcome to submit a PR, just make sure to measure the performance of DBSCAN with the original non-DenseMatrix vs the matrix converted to dense, so that we would not get surprising regressions.

[maximilian-gelbrecht]

ok, i dont have a lot of time this week, but i ll try to take a look at it next week.

[tlienart]

thanks to @oxinabox #193 will fix some of this too

[alyst]

Now dbscan() supports any AbstractMatrix, including the sparse ones.