Skip to content
Tutorials Center
My learning

Topics

Lesson 12 of 36 ยท Membership Functions and Operations

T-norms and S-norms

Learn the general families of fuzzy AND and OR operators, t-norms and s-norms: minimum, product, Lukasiewicz, drastic and Hamacher, with their axioms, duality and Python code.

In the last lesson AND was min and OR was max. These are the standard choices but not the only ones. Sometimes you want "tall AND fast" to be lower than either input alone (as with the product, where 0.7 * 0.6 = 0.42) so that every condition counts, not just the weakest. Fuzzy logic therefore defines families of operators that all behave sensibly as "AND" or "OR". The AND family is called t-norms (triangular norms) and the OR family s-norms or t-conorms.

You will meet the choice again in Fuzzy IF-THEN Rules and Mamdani Inference, where it decides how strongly a rule fires.

What makes a t-norm

A function T(a, b) taking two degrees and returning a degree is a t-norm if it satisfies four axioms:

  1. Commutative: T(a, b) = T(b, a)
  2. Associative: T(a, T(b, c)) = T(T(a, b), c), so three or more conditions can be combined in any grouping
  3. Monotone: if a <= c and b <= d then T(a, b) <= T(c, d) (more truth in, more truth out)
  4. Boundary: T(a, 1) = a (anding with "fully true" changes nothing)

A function S(a, b) is an s-norm (t-conorm) with the same first three axioms, but its neutral element is 0: S(a, 0) = a.

The two families are linked by a De Morgan-style duality through the standard complement: S(a, b) = 1 - T(1 - a, 1 - b). So each t-norm has a matching s-norm.

The common pairs

T-norm (AND) Formula Matching s-norm (OR) Formula
Minimum (Zadeh) min(a, b) Maximum max(a, b)
Algebraic product a * b Probabilistic sum a + b - a*b
Lukasiewicz (bounded difference) max(0, a + b - 1) Bounded sum min(1, a + b)
Drastic product b if a = 1, a if b = 1, else 0 Drastic sum b if a = 0, a if b = 0, else 1
Hamacher product a*b / (a + b - a*b) Hamacher sum (a + b - 2ab) / (1 - ab)
Python Five t-norms on the same inputs

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 = 0.8 and b = 0.6 the minimum gives 0.6, the product 0.48, the Lukasiewicz norm 0.4 and the drastic product 0, which shows the ordering. Every t-norm lies between the drastic product (the smallest) and the minimum (the largest):

TEXT
drastic <= Lukasiewicz <= product <= min

and dually for s-norms:

TEXT
max <= probabilistic sum <= bounded sum <= drastic sum

A smaller t-norm is a "stricter AND", a bigger s-norm is a "more generous OR".

flowchart LR
    D["Drastic<br/>(strictest AND)"] --> L["Lukasiewicz"] --> P["Product"] --> M["Minimum<br/>(most lenient AND)"]

T-norms are ordered from the strictest to the most lenient.

Duality in code

Given any t-norm, the matching s-norm comes for free from the duality formula. This is handy when you build your own library.

Python Deriving s-norms from t-norms

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.

The three derived ORs (0.70, 0.82, 1.00) match max, the probabilistic sum and the bounded sum exactly.

Combining many conditions

Associativity lets you fold a t-norm over any number of conditions. A rule such as "IF temperature is high AND pressure is high AND flow is low" has three antecedents; its firing strength is T(T(a, b), c). Here the choice has a visible effect: with the minimum, one weak condition sets the strength; with the product, every condition drags the strength down.

Python Firing strength with three conditions

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.

With eight conditions at 0.9, the minimum stays at 0.9, the product shrinks to about 0.43 and the Lukasiewicz norm falls to 0.2. That is why the minimum is the default in many systems: it does not punish a rule for having many reasonably true conditions. The product keeps more information and gives smoother outputs, and it is the usual choice in Sugeno systems and in neuro-fuzzy learning because it is differentiable.

Choosing an operator

  • Use min / max as the default: simple, fast, widely understood.
  • Use product / probabilistic sum when you want every input to influence the result, or need smooth derivatives.
  • Use Lukasiewicz / bounded sum in logic-oriented work, where degrees behave like "amounts" that add up.
  • Avoid the drastic operators in practice: they are discontinuous and mainly of theoretical interest.

Be consistent. Pick one AND/OR pair for the whole system, and keep it paired with its dual. Mixing a product AND with a max OR works in practice but loses De Morgan duality, so some equivalent rules will behave differently.

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. Which axiom says that anding with 1 changes nothing?

    The boundary condition says 1 is the neutral element of a t-norm.
  2. 2. What is the algebraic product t-norm of 0.5 and 0.6, and its matching s-norm (probabilistic sum)?

    0.5 * 0.6 = 0.30 and 0.5 + 0.6 - 0.30 = 0.80.
  3. 3. Which ordering of t-norms is correct, from smallest to largest?

    The drastic product is the strictest, the minimum the most lenient t-norm.
  4. 4. How is the s-norm dual to a t-norm T defined?

    This is the fuzzy De Morgan duality using the standard complement.
  5. 5. Why does associativity matter for fuzzy rules?

    A rule with three or more conditions needs T to be applied repeatedly, and associativity makes the result independent of the order.

Summary

  • T-norms generalise AND, s-norms (t-conorms) generalise OR; both are commutative, associative and monotone.
  • T-norms have neutral element 1, s-norms have neutral element 0.
  • Common pairs: min/max, product/probabilistic sum, Lukasiewicz/bounded sum, Hamacher, drastic.
  • Every t-norm lies between the drastic product and the minimum.
  • Duality S = 1 - T(1 - a, 1 - b) derives the OR for each AND.
  • Default to min/max; use product-based pairs for smoothness or learning.

Next: Fuzzy Relations and Composition