How to measure and tune quantum devices using machine learning – seminarium

Akademickie Forum Dyskusyjne Logos zaprasza na seminarium „How to measure and tune quantum devices using machine learning”, które poprowadzi prof. Andrew Briggs (Department of Materials, University of Oxford). Seminarium odbędzie się 16 października 2019 r. o godz. 9.45 w łączniku A3/A4 s. 105.


As the race to scale up reliable quantum computing accelerates, fault-tolerant error correction requires each logical qubit to be encoded in many physical qubits. A generic problem, common to all implementations, is device variability, whether in the gates of an ion trap or the electrostatic confinement of a solid state device. Electron spins in semiconducting devices offer a long-term platform for quantum computing, inspired by integrated circuits, with either the spin state or the relative spin alignment of two electrons representing the qubit. The elements of stabiliser codes (initialisation, one and two-qubit gates, single-shot readout) have been demonstrated, but a major obstacle to creating large circuits is variability due to trapped substrate charges. Tuning a single qubit requires searching a multi-dimensional gate voltage space, where device parameters vary non-monotonically and not always predictably with gate voltage. Tuning large multi-qubit circuits will require an automated approach. Advances in machine learning are becoming available for this purpose. Bayesian optimisation provides the basis for enabling the machine to decide what data to measure next in order to yield the greatest benefit in updating its knowledge of the device characteristics. This provides an automated pathway to converging in the shortest possible time on the configuration of parameters for the required performance. These methods are already being used to tune quantum dot devices, and they are applicable to a wide range of other platforms for quantum computing. I foresee that before too long, we shall be wondering how we ever managed without them!

Reference: Efficiently measuring a quantum device using machine learning, arXiv:1810.10042