Machine learning assisted quantum computing
Title: Machine learning assisted quantum computing
DNr: Berzelius-2024-222
Project Type: LiU Berzelius
Principal Investigator: Mats Granath <>
Affiliation: Göteborgs universitet
Duration: 2024-05-29 – 2024-12-01
Classification: 10302


The development of quantum computers is a major international effort, which involves research both at academic institutions and high-tech companies. Present day quantum computers are strongly limited by noise, with qubits loosing coherence quickly on the time scale of the operating times of circuit gates. We work on two main projects of relevance to the problem of qubit decoherence: 1) Quantum error correction (QEC). QEC is directed at the core problem of how to construct long-lived quantum states by distributing the information (quantum state) over many physical qubits. 2) Quantum circuit compilation (QCC). QCC deals with the problem of how to map a general (unitary) operation on a set of qubits onto actual hardware. The latter, for practical reasons, has a limited set of available quantum gates, and due to decoherence it is crucial to minimize the length (time) of any circuit. It turns out that both of these questions involve high-complexity optimization problems, amenable to classical machine learning. We use a range of deep machine learning algorithms, including graph neural networks (using Pytorch Geometric) and deep reinforcement learning (relying on JAX and Pytorch). The projects are part of and partly funded by the Wallenberg Centre for Quantum Technology (WACQT). Recent publications are listed in the Activity report for project NAISS 2023/23-302