Benefit of noise-encoded PQFM features for analyzing noisy financial observables

Ascertain whether and how noise-encoded transformed features produced by Heisenberg-ansatz PQFM circuits on IBM Quantum Heron hardware benefit the analysis and modeling of noisy financial observables for estimating RFQ fill probabilities, and characterize the conditions under which such features provide an advantage over classical features or noiseless quantum-simulated PQFM features.

Background

The authors observe that PQFM features generated on noisy quantum hardware lead to smoother, more normalized distributions and are associated with improved out-of-sample AUCs in fill probability estimation, yet they lack a clear understanding of why and when these noise-encoded features confer benefits in financial modeling.

This uncertainty persists despite tests showing similar uplifts across multiple IBM Heron devices and the inability of simulated noise models to reproduce the hardware-induced performance gains.