FAQ

Is SuperQuantX production software? No. It is a research toolkit. Everything in the showcase runs on local classical simulators.

Does it show quantum advantage? No, and it is built to tell you so. Every algorithm reports a classical reference: exact diagonalisation for VQE, brute force for QAOA and the portfolio agent, and tuned classical models for classifiers via compare_to_classical.

Which backends work? simulator, pennylane, qiskit, cirq, braket (local simulator) and tket are tested. D-Wave ocean handles QUBO/Ising problems only. Cloud hardware through those SDKs is untested. Azure Quantum, Rigetti Forest, cuQuantum and TensorFlow Quantum are planned, not implemented. Run superquantx list-backends to check your installation.

Why do all backends give the same accuracy? They compute the same exact statevector kernel. Identical results are a correctness check.

Why is the untuned quantum kernel so weak? A fidelity kernel with bandwidth 1 is often too narrow or too wide for the data. Tuning the bandwidth (the gamma argument of QuantumSVM) matters more than the choice of feature map; see arXiv:2610.42029.

Does the library configure logging? No. It only adds a NullHandler. Configure logging in your application.