# Which theories and concepts exist where one calculates with sets?

Recently I thought about concepts for calculating with sets instead of numbers. There you might have axioms like:

1. For every $a\in\mathbb{R}$ (or $a\in\mathbb{C}$) we identify the term $a$ with $\{a\}$.

2. For any operator $\circ$ we define $A\circ B := \{a\circ b : a\in A\land b\in B\}$.

3. For any function $f$ we define $f(A) := \{ f(a) : a\in A \}$. (More general: For a function $f(x_1,\ldots, x_n)$ we define $f(A_1,\ldots, A_n):= \{f(a_1,\ldots, a_n): a_1\in A_1 \land \dots \land a_n\in A_n \}$).

4. One has to find a good definition for $f^{-1}(A)$ which might be the inverse image of $A$.

((3.) is just the normal definition of the image and (2.) is a special case of (3.))

Now I am interested to learn about theories and concepts where one actually calculates with sets (similar to the above axioms).

After a while I found interval arithmetic. What theories or approaches do you know?

Because there will not be just one answer to my question, I will accept the answer with the most upvotes.

Update: The theories do not have to follow the above axioms. It's okay when they make there own definitions how a function shall act on sets. It is just important that you calculate with sets in the theory, concept or approach.

-
Are you sure that $x^2 = 4 \Rightarrow x = \{ 2, -2 \}$? Looks like $x$ could be $\{2\}$ or $\{-2\}$, according to your item 3. –  user31373 Jun 24 '12 at 18:55
You are right. I will delete my examples because they might be confusing... –  tampis Jun 24 '12 at 18:59
You could simply write $x\in \{2,-2\}$ instead of $x=\{2,-2\}$. This holds without changing any axioms or creating any new ones. For the second example $x\in [-1,1]$ will suffice. –  Andrew Salmon Jun 24 '12 at 19:01
Monoid of subsets of a group have been mentioned at MSE a few times, e.g. here. Proofwiki mentions some basic properties. –  Martin Sleziak Jun 27 '12 at 7:10
@MartinSleziak thx for the links ;-) –  tampis Jun 27 '12 at 7:12

I like Minkowski addition, aka vector addition. It is a basic operation in the geometry of convex sets. See: zonotopes & zonoids, Brunn-Minkowski inequality, polar sets... and here's a neat inequality for an arbitrary convex set $A\subset\mathbb R^n$: $$\mathrm{volume}\,(A-A)\le \binom{2n}{n}\mathrm{volume}\,(A)$$ with equality when $A$ is a simplex. (Due to Rogers and Shepard, see here)
The case $n=1$ isn't nearly as exciting.