Description
Recent advances in quantum computing have enabled mid-circuit measurements and feedforward, allowing real-time adjustments to quantum circuits based on earlier measurement outcomes. This capability is essential for a wide range of applications, such as measurement-based quantum computing, while loops, the ability to reset and reuse qubits during a computation, and error correction.
As powerful as this technology is, it has an Achillies heel: in practice quantum devices often suffer from significant levels of readout noise, wherein measurement outcomes are inaccurately identified and reported. Each incorrect measurement outcome reported by the device not only corrupts the measurement statistics (potentially used to compute observables) but also causes the subsequent execution of incorrectly chosen quantum operations. The effect is that the quantum state is fed into an incorrect branch of the program -- this can fundamentally change the program being evaluated. This complication has made such errors immune to all existing methods of error mitigation, which fail as they make no attempt to repair the broken correlations between the true measurement outcomes and feedforward quantum operations.
Here we give a protocol for mitigating mid-circuit measurement errors on circuits with feedforward, allowing us to recover the correct computational results in spite of these errors. We show our method applies to circuits with multiple layers of measure and feedforward opperations, incurrs no overhead beyond an increase in the sample complexity, and is robust against misscalibratin errors.
We tested our method on multiple IBM superconducting processors including the latest generation devices, showing experimentally that it reduces the impact of errors across a broad domain of circuits, improving fidelities of computed outcomes by up to 60%. Indicating the gaduate will be useful for developing and demonstrating algorithms on near and intermediate term devices.
| I am the presenting author | Yes |
|---|