Skip to content
Tutorials Center
My learning

Topics

Lesson 8 of 36 ยท Fuzzy Sets

Alpha-Cuts and the Extension Principle

Slice fuzzy sets into crisp alpha-cuts, do arithmetic with fuzzy numbers such as about 3 plus about 5, and apply ordinary functions to fuzzy sets with the extension principle.

How do you add "about 3" to "about 5"? How do you square a fuzzy number? Ordinary arithmetic works on exact numbers, so we need a way to carry it over to fuzzy sets. Two connected ideas solve this. Alpha-cuts slice a fuzzy set into a stack of ordinary crisp sets. The extension principle then lets us apply any normal function to a fuzzy set by working on those slices.

These tools also give us the "resolution identity", an elegant fact: a fuzzy set can be rebuilt entirely from crisp sets. Read Properties of Fuzzy Sets first if support and core are not yet familiar.

Alpha-cuts

Pick a level alpha between 0 and 1. The alpha-cut of a fuzzy set A is the crisp set of all elements whose membership is at least alpha:

TEXT
A_alpha = { x : mu_A(x) >= alpha }

The strong alpha-cut uses > instead of >=. As alpha rises, the cut shrinks. For alpha near 0 you get (almost) the support; for alpha = 1 you get the core. For a convex set each cut is a single interval.

flowchart TD
    A["Fuzzy set A"] --> C0["alpha = 0.2<br/>wide interval"]
    A --> C5["alpha = 0.5<br/>medium interval"]
    A --> C9["alpha = 0.9<br/>narrow interval"]
    C0 --> N["Nested: higher alpha, smaller set"]
    C5 --> N
    C9 --> N

Alpha-cuts are nested: a higher threshold gives a smaller crisp set.

Python Alpha-cuts of a triangular set

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

For a triangle the cut endpoints move linearly with alpha, which is why the sampled answers match the formula [a + alpha(b - a), c - alpha(c - b)].

The resolution identity

Every fuzzy set can be reconstructed from its alpha-cuts: the membership of a point x is the highest alpha for which x still lies in the cut. In symbols, mu(x) = sup { alpha : x in A_alpha }. This fact (often called the decomposition theorem) says that fuzzy sets are not an exotic species: they are stacks of crisp sets, and any operation defined on crisp sets can be extended to fuzzy sets by applying it to every slice and stacking the results.

Arithmetic with fuzzy numbers

A fuzzy number is a normal, convex fuzzy set on the real line with one peak, such as "about 3". Its alpha-cuts are intervals [lo, hi], so fuzzy arithmetic becomes interval arithmetic on each level:

  • addition: [a, b] + [c, d] = [a + c, b + d]
  • subtraction: [a, b] - [c, d] = [a - d, b - c]
  • multiplication of positive intervals: [a, b] * [c, d] = [a*c, b*d]

Do this for every alpha and stack the results to get the fuzzy answer.

Python Adding about 3 and about 5

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

At alpha = 0 the sum spans 6 to 11 (the support) and at alpha = 1 it collapses to the single point 8 = 3 + 5. For triangular numbers, adding means adding the three corner values, which is why triangular fuzzy numbers are popular in practice: the result stays triangular.

The extension principle

Now suppose f is an ordinary function, say f(x) = x * x, and A is a fuzzy set. What is f(A)? The extension principle answers: the membership of y in f(A) is the best (largest) membership among all x that f maps to y:

TEXT
mu_f(A)(y) = max { mu_A(x) : f(x) = y }

If no x maps to y, the membership is 0. When several x collide on the same y, we keep the highest degree: a value is as true as its most true source.

Python Squaring a discrete fuzzy set

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Look at the square: both -1 and 1 map to 1, with degrees 0.8 and 0.6, so the result has 1: 0.8. Squaring loses the sign, and the extension principle keeps the strongest claim. The shift A + 3 simply relocates the same degrees, because x + 3 never maps two inputs to one output.

For several variables the principle takes the minimum of the input degrees when pairs (x1, x2) collide on the same output, and then the maximum over all colliding pairs. That is what the alpha-cut method computes smoothly for continuous fuzzy numbers.

Why this matters later. The extension principle is what justifies fuzzy relations and their composition (Fuzzy Relations and Composition), and it is the theoretical basis for fuzzy arithmetic in risk analysis, scheduling and engineering with uncertain quantities.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. What is the alpha-cut at alpha = 0.5 of a fuzzy set?

    A_alpha = { x : mu(x) >= alpha }, a crisp set.
  2. 2. What happens to alpha-cuts as alpha increases?

    A higher threshold admits fewer elements, so the cuts are nested inside each other.
  3. 3. The alpha-cut of about 3 = (2, 3, 4) at 0.5 is [2.5, 3.5]. About 5 = (4, 5, 7) gives [4.5, 6]. What is the cut of their sum?

    Add the lower ends and the upper ends: 2.5 + 4.5 = 7 and 3.5 + 6 = 9.5.
  4. 4. Two inputs x1 and x2 are mapped to the same y by f. How does the extension principle combine their degrees?

    mu(y) = max of mu(x) over all x with f(x) = y.

Summary

  • An alpha-cut is the crisp set of elements with membership at least alpha; cuts are nested.
  • A fuzzy set can be rebuilt from its cuts (resolution identity).
  • Fuzzy-number arithmetic is interval arithmetic on each alpha level; triangular numbers add corner by corner.
  • The extension principle applies a function to a fuzzy set by taking the maximum membership over all sources of each output.
  • These ideas support fuzzy relations and fuzzy arithmetic.

Next: Triangular and Trapezoidal Membership Functions