Learning Pure Quantum States (Almost) Without Regret

Learn about quantum state tomography with minimal sample disturbance through this 19-minute conference talk from the Quantum Techniques in Machine Learning (QTML) 2025 conference. Explore the challenge of learning precise descriptions of quantum states through sequential measurements while minimally perturbing post-measurement sample states. Discover how regret is defined as cumulative disturbance across all samples and examine the balance between maximally informative measurement sequences and minimal regret measurements. Follow the presentation of a novel protocol for pure qubit states that achieves maximal precision while maintaining regret growth that scales only polylogarithmically with sample numbers. Understand the optimality proof for this scaling approach and gain insights into advanced quantum measurement strategies that preserve quantum information while enabling effective state characterization.