Simple scaling operations can change the monotone ordering according to size of the (dis)similarity coefficients and, hence, affect what clusters the analysis produces.
In the construction of the function spaces one could drop condition (3.7) and work with monotone functions only.
Section 2 deals with distributions having a monotone likelihood ratio.
All non-monotonic elements of the program are approximated by monotone constructs.
We do this by assigning indices to 0 monotone inductive operators.
For monotone and anti-monotone aggregates the truth value can be computed directly on the boundary multisets.
In section 4 we give an interpretation of the types of our systems as monotone operators on value sets.
In this logic two operators and are present, which express the least and greatest fixpoints of monotone operators on sets.
