TODO.md

List of possible additional features

Settings mechanism

Functionality in cnnclustering.settings currently not used.

Handling of meta information mappings

• Labels

Provide sanity checks for component combinations

Cleanup get_subset methods

Increase test coverage

• NeighboursGetter

DBSCAN clustering

Special fitter:

• Needs to only retrieve one neighbourlist
• No explicit similarity checker needed

Relaxed CommonNN clustering

Allow check of similarity criterion between all pairs of points removing the restriction that checked points are neighbours of each other.

Special fitter:

• Needs to loop over all (unassigned) points for checking instead of looping only over members of current neighbour list

Volume scaled CommonNN clustering

Special similarity checker:

• Requires not only two neighbour lists and a CommonNN cutoff but also pairwise distances

  • Different check signature Checker.check(input_data, na, nb, c, metric)
  • Special fitter which uses the specific check signature

• Restricted to euclidean distances metrics

• Input data needs to allow retrieval of neighbours and components

Clustering initialisation

Scan for non-sensical keywords that may have been passed (warning)

Children

defaultdict(
  lambda: Clustering(parent=self)
  )