Learning high-accuracy error decoding for quantum processors

Area of Science:

• Quantum computing
• Quantum error correction
• Machine learning

Background:

• Quantum error correction is crucial for building large-scale quantum computers.
• Quantum error-correction codes encode information redundantly across multiple qubits.
• Accurate decoding of noisy syndrome information is a key challenge.

Purpose of the Study:

• To develop a machine learning-based decoder for the surface code, a leading quantum error-correction code.
• To improve the accuracy of decoding noisy syndrome information for quantum computation.

Main Methods:

• Developed a recurrent, transformer-based neural network.
• Trained the network on simulated and real-world data from Google's Sycamore quantum processor.
• Utilized soft readouts and leakage information for enhanced decoding.

Main Results:

• The neural network decoder outperformed state-of-the-art decoders on real-world data for distance-3 and distance-5 surface codes.
• Maintained performance advantage on simulated data up to distance 11 with realistic noise.
• Demonstrated adaptation to unknown error distributions using experimental samples.

Conclusions:

• Machine learning can surpass human-designed algorithms in quantum error decoding.
• The developed decoder shows strong potential for practical application in quantum computers.
• This work highlights the power of data-driven approaches in advancing quantum technologies.