Usage: Hierarchical cannot be invoked directly from the command line

Wrapper for Hierarchical clustering

Brief Algorithm Description:

The hierarchical wrapper is a meta-wrapper which builds a cluster hierarchy by spawning off multiple runs of a given algorithm. The wrapper does not depend on a given algorithm being mccv_persistent, but it will help performance as many runs may be dispatched in parallel.

Data/parameter files are not created for each algorithms. Instead, its run() method is invoked directly

Required Parameters: (note: the list enclosed in the brackets are possible values each one of parameters can take )

Optional / Dependent Parameters:

hier_terminator = Terminator.<terminator function>

Provides a function that takes an ML_Algorithm and, based on internal criteria, return true or false depending on whether the algorithm's data satify some termination requirements. IF no hier_terminator parameter is specified, a default of clusterSizeTerminator(10) is used.

The built-in terminators are:

Always returns false. Effect is that clustering continues until the clustering algoruthm returns a single cluster per node

Terminate when the number of datapoints in a cluster is <= size

Terminates when the ratio of data points to number of dimensions falls below the specified ratio