This package implements the parameter free version of the DILCA distance. This approach allows to learn value-to-value distances between each pair of values for each attribute of the dataset. The distance between two values is computed indirectly based on the their distribution w.r.t. a set of related attributes (the context) carefully chosen.