Q-FIE
An interactive academic demo of fuzzy inference with classically simulated quantum circuits.
At a glance
- Illustrative model in the application
7 symptoms x 5 urgency categories- Backend tests passing in the current checkout
72- Analytic threshold in the example model
0.75
Problem
The research question is how two fuzzy implication rules change the output of a compositional inference pipeline. A formula alone hides the intermediate relation matrix and the point at which a classification changes.
Approach
Q-FIE’s Q-CRISP application lets a visitor adjust seven illustrative symptom values, compare Reichenbach and quantum Lukasiewicz implication, and inspect the resulting matrix and five urgency degrees. It places the immediate closed-form calculation beside an on-demand Qiskit Aer simulation of the corresponding circuits.
Architecture
The Python core holds the constants, closed-form implication and aggregation functions, circuit builders and simulator validation; FastAPI exposes the analytic calculation and asynchronous circuit jobs; a React frontend displays sliders, a relation heatmap, urgency bars and the comparison; SQLite stores demonstration patient records and their computed classification snapshots, while the detail view recomputes the full pipeline from the stored inputs.
Measured results
The configured model has 7 symptom values and 5 categories, producing a 7 x 5 relation
matrix. In the README’s worked third example, the Reichenbach urgency degree for the
“Very Urgent” category is 0.7807, while the quantum Lukasiewicz value is 0.7498; they
fall on opposite sides of the illustrative 0.75 threshold. The current checkout passes
72 backend tests, including formula, parity, API and patient-record tests
(python -m pytest -q backend/app/tests).
Engineering decisions
The analytic path keeps UI feedback immediate, while the circuit path runs as a separate job because simulation is more expensive. Putting the equations in one backend core gives the API and saved-record workflow the same calculation. Exposing both implication rules and intermediate values makes the classification change inspectable.
Limitations
This is an academic, illustrative demo, not a clinical decision tool. Qiskit Aer runs the circuits on a classical computer; there is no quantum hardware result or clinical validation. The README’s “42 tests” statement predates the current 72-test suite.