Work

The Mathematics of Quantum Computing Advantage

Year: 2026
Type: preprint
Abstract: When does a more expensive quantum measurement reduce the total cost of distinguishing two known states? We study sequential experiments in which each observation has a price and the experimenter can decide what to measure next or when to stop. For a family of qubit experiments with two readouts, we determine the smallest worst-case expected cost at each prescribed error tolerance. Randomizing between at most two stopping thresholds attains this cost. We characterize which pairs of qubit states realize the required observation laws. For an explicit pair of full-rank, noncommuting states, the second readout can carry a price premium of just under twenty percent and still reduce total cost at sufficiently high confidence. At a premium of twenty percent or more, the first readout alone is optimal at every confidence level. Exact rational calculations and an independent dynamic programming calculation certify these examples. We then consider a calibrated binary phase oracle subject to dephasing. Each experimental block incurs a fixed preparation cost and a cost proportional to its depth. The experimenter can adapt the depth and measurement using earlier outcomes, while retaining only classical information between blocks. We give upper and lower bounds of the same order for the optimal expected cost. They show how the phase separation, coherence time, and available circuit depth jointly limit the useful duration of a block. In the noiseless model, we determine the exact cost of discrimination with zero error. At high confidence, measured relative entropy determines how much evidence a measurement supplies per unit cost. For finite catalogues of measurements and several competing hypotheses, we show how finite decision programs approximate the minimax cost. An explicit bound controls the error caused by truncating the experiment. When each available measurement uniformly separates every pair of hypotheses, this error decreases exponentially with the decision horizon.
Language: English
Open Access status: green
OpenAlex ID: W7236504316
Cites:
Cited by:
Related to: 10