Scalable Neural Decoder for Topological Surface Codes.

Area of Science:

• Quantum Computing
• Quantum Error Correction
• Machine Learning

Background:

• Noisy Intermediate-Scale Quantum (NISQ) devices offer practical quantum computing potential.
• Scaling quantum circuits requires fast and efficient quantum error correction algorithms.
• Current decoders face limitations in scalability and speed for large quantum systems.

Purpose of the Study:

• To develop a scalable, neural network-based decoder for quantum error correction.
• To improve decoding times and accuracy compared to existing methods.
• To enable larger quantum circuits beyond the NISQ era.

Main Methods:

• A neural network decoder utilizing a sliding preprocessing window, inspired by convolutional neural networks.
• Application to stabilizer codes with depolarizing noise and syndrome measurement errors.
• Comparison with Union-Find and Minimum Weight Perfect Matching decoders.

Main Results:

• The decoder scales to tens of thousands of qubits.
• Achieves faster decoding times than the Union-Find decoder for error rates down to 1%.
• Reduces error rates by up to 2 orders of magnitude.
• Increases the error threshold by up to 15% compared to conventional algorithms, even with measurement errors.

Conclusions:

• The neural network decoder offers a scalable and efficient solution for quantum error correction.
• This approach effectively reduces error rates and improves error thresholds.
• Machine learning-assisted quantum error correction is crucial for advancing quantum computing beyond NISQ devices.