Quantum computing has long been plagued by the challenge of calibration drift. Superconducting qubits, a crucial component in many quantum systems, require precise calibration to function correctly. However, this traditional process can’t be performed while calculations are being done, leading to errors and reduced performance.
Google’s researchers have now developed a solution that uses error correction data to continuously recalibrate quantum processors. This breakthrough has the potential to improve the performance and reliability of quantum computers, enabling them to perform more complex calculations.
The Challenge of Calibration Drift
Calibration is crucial for devices like superconducting qubits, which are controlled by pulses of microwave photons. These pulses can drift from their initial settings due to random factors like hardware heating up as it’s used. This can be a problem for complicated algorithms that could crack current encryption.
Currently, if the system shows signs of drifting away from calibration, Google stops the computations and recalibrates. However, this is not an option partway through a calculation. The researchers point out that some errors detected by error correction are caused by calibration failures, making it challenging to tell them apart from random errors.
Reinforcement Learning Solution
The team’s solution uses reinforcement learning to identify the most effective adjustments to control parameters during computations. They deliberately apply small perturbations to all control parameters and score their effectiveness at limiting errors. This allows the system to infer how adjusting these parameters can minimize certain errors, making adjustments in real-time.
Experimental Results
The researchers put their system in charge of two logical qubits hosted on a calibrated system. They used different error correction schemes (a surface code and a color code) and set them in a specific state. The error-correction system was then used with and without reinforcement-learning-driven corrections, showing a 20 percent increase in the ability to detect and correct errors in the logical qubits.
Limitations and Future Work
The approach only works if the drift keeps the system reasonably close to its initial state. If it’s significantly different, the corrections may not be effective. To address this, the researchers suggest constantly re-evaluating the effectiveness of different changes. However, this raises a problem: You can’t simply randomize all potential control configurations in the middle of a calculation.
The team performed many simulations with a small error-corrected qubit and showed that the trade-off worked out, provided drift was slow enough. They also demonstrated it could work in real-time with a large error-corrected qubit, where the reinforcement learning system had control over 40,000 parameters.
Conclusion
Google’s researchers have developed a method to continuously recalibrate quantum processors using error correction data. This breakthrough has the potential to improve the performance and reliability of quantum computers, enabling them to perform more complex calculations.
Source: Original article