MISIS: Neural network for finding and correcting errors in quantum computing

The main articles are:

• Neural networks (neural networks)
• Quantum computer and quantum communication
• Quantum computers and networks in Russia

2025: Russian scientists train neural network to correct errors of quantum computers

Researchers at MISIS University have created a neural network-based system that learns to find and correct errors in quantum computing. The development combines the advantages of intelligent and classical algorithms, so it more effectively recognizes errors that occur when the number of qubits increases - the "building blocks" of quantum processors. The university announced this on February 14, 2025.

Quantum computers, unlike conventional computers, use qubits - elements capable of performing complex calculations faster than classical processors. However, qubits are extremely sensitive to interference: even the slightest external impact can distort data. To increase the reliability of quantum computing, scientists at MISIS University have presented an algorithm based on recurrent neural networks that learns to detect errors.

Modern devices make mistakes largely due to the interaction of a quantum system with its environment. At the same time, even small errors are critical for scale calculations, since the distortion of the result accumulates with each operation. Improving accuracy is one of the key tasks in the development of quantum technologies, "said Alexey Fedorov, director of the Institute of Physics and Quantum Engineering at NITU MISIS.

The method is based on the architecture of recurrent neural networks for analyzing time series of data obtained during periodic measurement of auxiliary qubits. This feature allows the algorithm to work with different correction codes. The researchers tested the algorithm on a family of cyclic correction codes, given the topological features of a superconducting quantum processor. The results of the study are published in the journal Physical Review A (Q1).

The main quality of development is the ability to learn from data received from a specific device. This is especially important in conditions where the nature of errors differs from theoretically assumed models. In addition, the proposed decoding algorithm does not depend on the specific correction code, which makes it universal and easily scalable, - said the author of the study Ilya Simakov, engineer of the scientific project of the laboratory of superconducting quantum technologies NUST MISIS, researcher at the Russian Quantum Center.