MW Maida Wang
Maida Wang
portrait.jpg
PhD Student · University College London

Maida Wang

PhD Researcher at UCL working with Prof. Peter V. Coveney, specialising in Quantum Machine Learning and hybrid quantum–classical AI. I focus on developing algorithms that bridge the gap between the NISQ and FTQC eras to achieve practical quantum advantage. Passionate about translating deep-tech research into scalable solutions that drive the commercialisation of quantum technologies.

  • GroupCentre for Computational Science
  • SupervisorProf. Peter V. Coveney
  • 2nd SupervisorProf. Dan Browne
  • Funded byUCL Dean's Scholarship
  • Based inLondon, UK
Email Google Scholar UCL Profile CV
scroll
01

About

I work across the theoretical and experimental sides of quantum computing, and I am equally interested in both. On the theory side, my focus is the expressivity and generalisation of quantum machine learning models: covering-number arguments, effective dimension and active subspaces of parameterised quantum circuits, and the Fisher-information / gradient-covariance geometry induced by Lie-algebraic parameterisations. I also work on the information-theoretic side of QML. Recent work of mine proves a two-stage practical quantum advantage for Quantum-Informed Machine Learning, using Bell-type joint-copy measurements to extract post hoc Pauli statistics with copy complexity independent of qubit count, against an Ω(2kq) lower bound for any adaptive single-copy protocol.

On the experimental side, I have run hardware experiments on both superconducting processors (most recently IQM-Garnet) and photonic quantum devices. The work centres on parameter-efficient circuits, careful encoding schemes, and measurement protocols that survive realistic noise on near-term hardware.

Before UCL I did my BSc and MSc at Wuhan University (武汉大学), and before that I went to Changsha No.1 High School (长沙市第一中学). Outside of research I take long walks, take care of a Shiba Inu, and read more physics history than is good for me.

02

Research Interests

⟨ψ|

QML Theory

Expressivity bounds via covering numbers, effective dimension, and active subspaces of PQCs. Generalisation derivations under Lie-algebraic parameterisations.

|0⟩

Quantum Hardware

I run experiments on superconducting (IQM-Garnet) and photonic devices. The focus is shallow circuits, careful encoding, and Bell-type measurement protocols that hold up under realistic hardware noise.

∇·

Quantum-Informed PDEs

Quantum priors as a structural regulariser for chaotic systems such as Kuramoto–Sivashinsky and 2D Kolmogorov flow, with weather-scale forecasting as the long-term target.

03

News

  1. Award Received the 2026 Research Excellence Award from the UCL Faculty of Mathematical & Physical Sciences.
  2. New theory preprint (“Foundations of Practical Quantum Advantage in Quantum-Informed Machine Learning for Predicting Chaos”) posted on arXiv with Xiao Xue, Minh Chung, and Peter V. Coveney. arXiv →
  3. Press IQM Quantum Computers published an interview with me on achieving quantum advantage before the fault-tolerant era (“Why practical quantum computing can't wait for fault tolerance”). IQM Blog →
  4. Press Roger Highfield wrote about the QIML work on the Science Museum blog (“From turbulence to treatments: quantum computing boosts AI and medicine”). Science Museum →
  5. Press UCL News featured the QIML work (“Quantum computer improves AI predictions”). UCL News →
  6. Press LRZ (Leibniz Supercomputing Centre) featured the work in its newsroom (“Quantencomputer verbessern die Vorhersagen von KI”). LRZ →
  7. Paper “Quantum-Informed Machine Learning for Predicting Spatiotemporal Chaos with Practical Quantum Advantage” published in Science Advances. DOI →
  8. Preprint “Uni-Flow: a unified autoregressive-diffusion model for complex multiscale flows” released on arXiv with Xiao Xue and collaborators. arXiv →
  9. New preprint: “Q-BAR: Blogger Anomaly Recognition via Quantum-enhanced Manifold Learning.” arXiv →
  10. “Parameter-Efficient Quantum Anomaly Detection Method on a Superconducting Quantum Processor” published in Physical Review Research. DOI →
  11. Started Teaching Assistant roles at UCL: Machine Learning (Department of Computer Science) and Mathematical Modelling (Department of Engineering).
  12. Teaching Assistant at the UCL Department of Information Studies (MySQL).
  13. Started my PhD at UCL Centre for Computational Science under Prof. Peter V. Coveney and Prof. Dan Browne.
  14. My first paper as first author (“A Quantum-Classical Hybrid Solution for Deep Anomaly Detection”) appeared in Entropy.
04

Publications

* denotes corresponding author · † equal contribution. Full list on Google Scholar.

Thy2026

Foundations of Practical Quantum Advantage in Quantum-Informed Machine Learning for Predicting Chaos

Maida Wang, Xiao Xue, Minh Chung, Peter V. Coveney*

arXiv:2606.13422 · 2026

arXiv PDF DOI
SA2026

Quantum-Informed Machine Learning for Predicting Spatiotemporal Chaos with Practical Quantum Advantage

Maida Wang†, Xiao Xue†, Mingyang Gao, Peter V. Coveney*

Science Advances, 12(16), eaec5049 · 2026

DOI arXiv UCL News Science Museum LRZ IQM Blog
PRR2025

Parameter-Efficient Quantum Anomaly Detection Method on a Superconducting Quantum Processor

Maida Wang, Jinyang Jiang, Peter V. Coveney*

Physical Review Research, 7, 043094 · 2025

DOI arXiv
arXiv2026

Uni-Flow: A Unified Autoregressive-Diffusion Model for Complex Multiscale Flows

X. Xue, T. Yang, M. Gao, L. Pan, Maida Wang, K. Zhu, S. Wang, J. Li, M. F. P. ten Eikelder, Peter V. Coveney

arXiv:2602.15592 · 2026

arXiv PDF
arXiv2025

Q-BAR: Blogger Anomaly Recognition via Quantum-enhanced Manifold Learning

Maida Wang, P. Jiang

arXiv:2512.11071 · 2025

arXiv
Ent2023

A Quantum-Classical Hybrid Solution for Deep Anomaly Detection

Maida Wang, Anqi Huang*, Yong Liu, Xiang Yi, Junjie Wu, Songhui Wang

Entropy, 25(3), 427 · 2023

DOI MDPI
Conf2020

Using the Ternary Closed-loop Model to Research on Pterosaurs' Flight Capability

Maida Wang, Zhongxiang Zhang*

Proc. ICMAI · 2020

05

Projects

My PhD work, “Towards Practical Advantage with Quantum Machine Learning,” is organised around three threads. 2nd-Year Review poster (PDF) →

Experiment · QIML

Quantum-Informed Machine Learning for Chaotic PDEs

A hybrid quantum–classical framework for long-horizon prediction of chaotic dynamics. A small quantum generator (the Q-Prior) is trained on the IQM-Garnet superconducting chip and paired with a transformer-based classical predictor. Validated on Kuramoto–Sivashinsky, 2D Kolmogorov flow, and turbulent channel flows; achieves stable forecasts and a ~103:1 memory advantage over raw datasets. Published in Science Advances (2026), in collaboration with NVIDIA, IQM, and LRZ.

UCLNVIDIAIQMLRZ
Experiment · PEQAD

Parameter-Efficient Quantum Anomaly Detection

A quantum anomaly-detection framework optimised for shallow circuits and few measurements, deployed on real superconducting hardware and benchmarked on photonic devices. Combines amplitude encoding with a deep one-class objective to beat classical baselines in the small-sample regime. Published in Physical Review Research (2025).

UCLSuperconducting QPUPhotonic QPU
Theory · Foundations

Foundations of Practical Quantum Advantage in QIML

A theory paper formalising a practical quantum advantage mechanism for QIML on chaotic systems. We extend the single-site Q-Prior to a k-indexed family of higher-order quantum statistical priors on nq = kq qubits, and prove a two-stage advantage: (i) representation, where entanglement compactly stores non-factorisable spatial correlations of the invariant measure; and (ii) extraction, where Bell measurements on two copies estimate any post hoc Pauli functional with a copy-pair count independent of nq, while adaptive single-copy protocols require Ω(2nq) copies. Two case studies instantiate the mechanism: a turbulent channel-flow study that recovers a named non-diagonal correlator of the invariant measure, and a ECMWF ERA5 medium-range weather forecasting workflow inside a Koopman model (+10–39% ACC on Z500 over 48–240h lead times, long-horizon rollouts stable against mean-field collapse).

UCL CCSNVIDIALRZarXiv 2606.13422
Theory · Ongoing

Expressivity, Generalisation & Bell-Type Measurement

A broader theory thread on the learning-theoretic foundations of QML: expressivity bounds via covering numbers, effective dimension and active subspaces of PQCs; generalisation derivations under Lie-algebraic parameterisations and Fisher-information / gradient-covariance geometry; and the structural role of Bell-type measurement and non-classical correlations in QML protocols. Funded through EPSRC EP/W007711/1 + SEAVEA.

EPSRC EP/W007711/1SEAVEA
06

Education & Experience

  1. 2024.09 — present

    PhD in Quantum Computing and Machine Learning

    University College London (UCL), Centre for Computational Science.
    Supervised by Prof. Peter V. Coveney and Prof. Dan Browne.

  2. 2021.09 — 2024.06

    MSc in Applied Mathematics

    Wuhan University (武汉大学), Wuhan, China.

  3. 2017.09 — 2021.06

    BSc in Mathematics and Applied Mathematics

    Wuhan University (武汉大学), Wuhan, China.

  4. — 2017

    Changsha No.1 High School (长沙市第一中学)

    Changsha, Hunan, China.

  5. Industry · 2022 — 2024

    Industry Roles

    2086 Laboratory, China Electronics Corp. · Advanced Computing Engineer (quantum ML algorithms) · 2022.11 – 2024.06
    Baidu · Algorithm Engineer, Baidu Quantum Platform · 2022.08 – 2022.11

  6. Teaching · 2025 — 2026

    UCL Teaching Assistant

    Department of Computer Science, Machine Learning · 2025.09 – 2026.03
    Department of Engineering, Mathematical Modelling · 2025.09 – 2026.03
    Department of Information Studies, MySQL · 2025.01 – 2025.05

07

Get in Touch

I'm always happy to talk about quantum machine learning, chaotic dynamics, or anything sitting awkwardly between physics, statistics and computer science. The fastest way to reach me is email. For the academic record, my Google Scholar and UCL profile are the most up-to-date.