ILP Part 74 — Absolute value for real numbers

This is the seventieth fourth part of the ILP series. For your convenience you can find other parts in the table of contents in Part 1 – Boolean algebra

The original formula uses comparisons:

where should be an absolute value of . We know that is at least and we use the formula above to enforce the equality. Effectively, we want to say that is either equal to or

We know that this is very dangerous as it implicitly requires to be at least . However, we don’t need to compare numbers precisely. What we need is a formula which will not allow for to be greater than . Let’s introduce this unreliable comparison:

if then must be one. However, if then is already non-positive so can be either zero or one.

However, important part for us is that if then will be one. So now we can do two comparisons:

If then both comparisons would yield true which in turn would violate the constraint. Since we know that is at least then there is no other way as to set . This works because solver can choose any result for unreliable comparisons.

This can be used for integer values as well. We can effectively calculate absolute value without any comparisons (which break the model for real numbers). Same goes for maximum and minimum, we can calculate these in a safe manner.