Adds a few new options to KBMOD clustering and significantly updates the documentation. New options include:
The ability not to scale the position clusterings (these are encoded as _unscaled clustering types)
The ability to cluster on a combination of start and end positions
As a general cleanup, the various positional algorithms were combined into a single predicted position clustering algorithm that takes different times. For example the ClusterPositionFilter was replaced by using ClusterPredictionFilter with time=0 and ClusterMidPosFilter by ClusterPredictionFilter with the median time. This allows for a single (better tested) class and the easier ability to add new functions (like turning off scaling).
Performed a few additional small clean ups:
More informative names for the clustering algorithm strings (for logging).
The best point (according to LH) is chosen from each cluster. This was already happening because results were sorted by LH first, but this makes it explicit in case the ordering of results is changed somewhere else.
Adds a few new options to KBMOD clustering and significantly updates the documentation. New options include:
_unscaledclustering types)As a general cleanup, the various positional algorithms were combined into a single predicted position clustering algorithm that takes different times. For example the
ClusterPositionFilterwas replaced by usingClusterPredictionFilterwith time=0 andClusterMidPosFilterbyClusterPredictionFilterwith the median time. This allows for a single (better tested) class and the easier ability to add new functions (like turning off scaling).Performed a few additional small clean ups:
Closes #539