MarketMind-Q: Compiler-aware finance QML benchmark

[Taz33m]

Submission

MarketMind-Q is a compiler-aware finance QML benchmark. It compares classical ML baselines against a Qiskit quantum-kernel SVM, then uses qBraid to test whether the quantum-kernel estimation workload preserves kernel probabilities after compilation across QASM2/Qiskit and Cirq execution paths.

This PR includes the full submission source tree under submissions/marketmind-q/, including Python source, configs, frozen benchmark data, generated results, tests, notebook, web dashboard source, documentation, and slides.

Links

• Code repository: https://github.com/Taz33m/mktmind-qtm (also put in second commit to this PR)
• Live dashboard: https://taz33m.github.io/mktmind-qtm/
• Presentation: https://1drv.ms/p/c/03e11cee1c77546a/IQDw_VDqkbWrQ5pN5H4g-CxQAVg0A8jGMhH_s51hpA6pb20

qBraid Challenge Fit

• Algorithm: quantum-kernel SVM for short-horizon finance classification.
• Source representation: Qiskit QuantumCircuit compute-uncompute kernel circuits built from zz_feature_map.
• qBraid usage: qbraid.transpile(...) and ConversionGraph().
• Compilation strategies: qasm2_roundtrip and cirq_direct.
• Execution environments: qiskit_statevector, qiskit_shots_1024, cirq_statevector, and cirq_shots_1024.
• Output-quality metrics: source/compiled zero-state probability, absolute probability error, Hellinger distance, optional qBraid allclose.
• Compiled-resource metrics: depth, qubits, two-qubit gates, measurement count, serialized size, transpile seconds, execution seconds, shots.

Conclusion

The best exact preservation path in the submitted run is qasm2_roundtrip / qiskit_statevector, with mean absolute probability error 0.0000, mean depth 21, and mean two-qubit gate count 9. Cirq direct also preserved statevector probabilities to near numerical precision with a different resource profile.