Bandits roaming Hilbert space

Bandits roaming Hilbert space

🎙 Josep Lumbreras Zarapico 👥 8K 📅 January 20, 2026 ⏱ 33 min 👁 109 📄 original study 🧭 2026-08-15
Available in: English (current) Français

Keywords

multi-armed banditquantum stateregretexploration-exploitationquantum battery

Summary

This seminar from the Centre for Quantum Technologies introduces the multi-armed bandit problem, a classic reinforcement learning challenge balancing exploration and exploitation. The speaker, Josep Lumbreras Zarapico, extends this to a quantum setting where the environment is an unknown quantum state, and actions are observables to measure. The goal is to minimize cumulative regret, defined as the difference between the optimal expected reward and the obtained reward. The talk covers the formulation of the quantum bandit problem, its connection to state disturbance, and its application to quantum battery charging. The speaker presents algorithms based on Optimism in the Face of Uncertainty (OFU) and adaptive confidence regions, achieving logarithmic regret for pure qubit states, which is optimal. The presentation includes technical details on weighted least squares estimators and median of means techniques. The talk concludes with a discussion of practical applications, particularly optimizing energy transfer in quantum batteries.

147 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides a clear and rigorous introduction to quantum bandit problems, building from classical bandits to the quantum extension. The speaker effectively motivates the problem with practical applications like state disturbance and quantum battery charging. The argumentation is solid, with mathematical derivations and references to known results in bandit theory. The presentation of algorithms and their optimality is convincing, though some technical steps are summarized. The speaker acknowledges open questions and limitations, enhancing the credibility of the work.

Scientific Rigor, Source Quality, Title Accuracy

The talk demonstrates scientific rigor through its structured presentation and reliance on established bandit theory. The speaker references classical bandit literature and information-theoretic bounds. However, the description lacks explicit citations or links to specific papers, limiting verifiability. The title is appropriate and engaging, accurately reflecting the content. The seminar format suggests a high level of expertise, and the speaker’s affiliation with a reputable quantum research center adds credibility.

162 words

Title / Content Match

The title 'Bandits roaming Hilbert space' is catchy and accurately reflects the content, which explores multi-armed bandit problems in quantum settings.

Quality & Reliability

8/10

The talk presents original research in quantum bandit problems, with rigorous mathematical derivations and references to established literature. The speaker is from a recognized quantum research center. However, the presentation is a seminar and lacks peer-reviewed publication details in the description.

Key Moments

Cited Sources

  • Centre for Quantum Technologies — The talk is part of the Quantum Lunch Seminar Series at CQT, National University of Singapore.

Concurring Sources

  • Multi-armed bandit — The classical bandit problem is well-documented and provides the foundation for the quantum extension.
  • Quantum tomography — The talk compares the quantum bandit approach to standard tomography, highlighting differences in objectives.

Dissenting Sources

  • Quantum battery — The concept of quantum batteries is relatively new and may have varying definitions; the talk's specific model may not align with all literature.

Contribution & Novelties

The talk presents original research on quantum multi-armed bandits, showing that for pure qubit states, logarithmic regret is achievable, breaking the classical square-root barrier. This is a novel contribution to the field of quantum decision-making. The application to quantum battery charging is also innovative, demonstrating practical relevance.

Pour aller plus loin :

78 words

Radar Profile

The radar profile shows high scores in technical level and information quality, with slightly lower but still strong scores in quantity and reliability. This indicates a technically dense and reliable presentation, suitable for an expert audience.

Reliability 8/10