SuperconducTED
Fuzzy-logic noise modeling for quantum simulation
Last updated: May 29, 2026
Status · Active research · targeting IEEE QCE 2026
Role · Lead & Architect
Overview
Quantum hardware is noisy, and that noise drifts. IBM recalibrates its machines constantly, so a noise model fit to one day's calibration data is often stale by the next. SuperconducTED asks a different question: instead of fitting noise to a single snapshot, can a model capture noise in a way that transfers across calibration cycles? The approach uses fuzzy inference to turn raw calibration data into ensemble-based noise models for Qiskit Aer simulations of IBM backends.
The Problem
Standard noise models in the Kraus/Lindblad formalism are precise but brittle. They describe one device state at one moment. Because IBM's calibration values shift hour to hour, a model that is accurate on Monday can mispredict by Wednesday. The research target is a noise representation whose predictions stay useful as the underlying hardware drifts, validated against a published 2026 reference benchmark.
The Approach · Why Fuzzy
SuperconducTED treats calibration parameters (T1, T2, gate and readout errors) as fuzzy quantities rather than exact scalars. It uses Takagi-Sugeno-Kang inference with interval type-2 fuzzy sets, which answer with an interval rather than a single number. The intuition: a type-1 system says the membership of T1 = 142 µs is 0.85; a type-2 system says it is somewhere in [0.55, 0.85], because we are also unsure how unsure to be. T1 is the noise channel; type-2 is how much uncertainty we are allowed to carry about it. This complements the Kraus/Lindblad formalism rather than replacing it: fuzzy inference acts as a calibration-advisor layer that produces the parameters the simulator consumes.
The Pipeline
A five-stage flow turns raw calibration data into a validated noise model: calibration ingestion, fuzzification, inference, ensemble aggregation, and Qiskit Aer integration. Each stage is documented and version-controlled, with key design choices captured as architecture decision records.
The Calibration-Data Engine
Transferability can only be tested against real drift, so the project collects it continuously. A GitHub Actions cron polls an IBM backend hourly and commits each snapshot to a dedicated data branch, accumulating a time series of how the hardware actually changes. That data set is what lets the team measure whether a model trained on one window still predicts the next.
Status
Active research. The calibration engine is running and accumulating snapshots; the inference and ensemble-aggregation layers are implemented and under iterative review. Results are being prepared toward an IEEE QCE 2026 submission. [TODO: add concrete transferability results once citable.]
Team
A five-person team plus a faculty advisor at TED University. Mert leads the architecture, the decision record, and the calibration pipeline.
Tech Stack
Qiskit · Qiskit Aer · IBM Quantum · scikit-fuzzy · NumPy · SciPy · pytest · GitHub Actions · Python
Links
GitHub organization → github.com/SuperconducTED