Statistical Bias in D-Wave Qubits
Thomas Krauss, Alexander Giffen, Phillip Truppelli, Alan J. Michaels · Journal of Physics Conference Series · 2021
Abstract Recent access to commercially available quantum computing capabilities offers the potential for testing a host of new algorithms that take advantage of fundamentally different processing approaches. One such computing resource is D-Wave’s quantum annealing computer, which employs an energy minimization-based computing mechanism to find solutions to problems stated in an Ising formulation. These solutions, however, must be interpreted in terms of Monte Carlo-style probabilistic occurrences and histograms of candidate solutions that presume the underlying statistics of the qubits are “fair.” This paper quantifies the underlying fairness of individual qubits on a single D-Wave machine, showing that there is a measurable bias in the resulting solution as a function of which qubits are used in the calculation. Further, an example problem is demonstrated, whereby the same problem formulation yields distinctly different solution probabilities by changing which physical qubits are employed.