Skip to main content

Ferrifuzz

Built for uncertainty. Engineered for speed.

Ferrifuzz is a fuzzy logic library with a pure‑Rust core and a Python package on top. Both expose the same engine, so you can prototype in a notebook and ship the same model in Rust. It was written with fuzzy‑systems courses in mind: rules are plain English, results are reproducible, and mistakes produce error messages that tell you what went wrong.

pip install ferrifuzz
import ferrifuzz as ff

service = ff.Variable("service", 0, 10)
service["poor"] = ff.mf.trapezoid(0, 0, 2, 5)
service["good"] = ff.mf.triangle(2, 5, 8)
service["excellent"] = ff.mf.trapezoid(5, 8, 10, 10)

tip = ff.Variable("tip", 0, 30)
tip["low"] = ff.mf.triangle(0, 5, 10)
tip["medium"] = ff.mf.triangle(10, 15, 20)
tip["high"] = ff.mf.triangle(20, 25, 30)

system = ff.FuzzySystem()
system.add_input(service)
system.add_output(tip)
system += ff.Rule.from_string("IF service IS poor THEN tip IS low")
system += ff.Rule.from_string("IF service IS good THEN tip IS medium")
system += ff.Rule.from_string("IF service IS excellent THEN tip IS high")

print(system({"service": 7.5})) # {'tip': 22.6079...}

What you get​

  • Membership functions: triangle, trapezoid, gaussian, generalized bell, sigmoid, singleton, and interval type‑2 sets built from any two of them.
  • Linguistic variables with dict‑style terms, hedges (very, extremely, somewhat, ...) and NOT.
  • Fuzzy operators: 6 t‑norms, 6 t‑conorms, 3 complements.
  • Rules as strings: IF temp IS NOT very cold OR humidity IS high THEN fan IS fast AND alarm IS on, with parentheses and rule weights.
  • Inference: Mamdani, Sugeno/TSK (constant or linear consequents) and interval type‑2 (Karnik–Mendel, center‑of‑sets, height type reduction). Six defuzzification methods.
  • Fuzzy time series forecasting (Chen 1996, first and high order) with RMSE/MAE/MAPE metrics.
  • Batch inference over NumPy columns, optional matplotlib plots, and a VS Code extension that completes, checks and highlights rule strings.

The pipeline at a glance​

crisp input ──▶ fuzzify ──▶ fire rules ──▶ aggregate ──▶ defuzzify ──▶ crisp output
(μ per (AND / OR (max per (centroid,
term) t-norms) output term) bisector, …)

Where to next​

I want to…Go to
Install the packageInstallation
Build and run a system in five minutesYour first system
Learn the theory behind the codeConcepts
Look up what rule strings can containRule syntax
Work through a full course‑style projectTutorials
Look up a class or methodPython API · Rust API
Fix an errorReading error messages · FAQ
Update code written for 0.1Migrating from 0.1
Cite ferrifuzz in a report or thesisCite & license