Learning from Quantum Experiments
summary
Recent results show a practical protocol that uses random Clifford sampling and simple postprocessing (median‑of‑means on stored classical shadows) to predict many observables with about O(B log M / ε^2) measurements, often independent of system size for relevant families, though computational costs remain unclear. A separate rigorous result proves an exponential separation (2^n vs n) between incoherent (classical‑memory) and coherent (quantum‑memory) experiments for certain Pauli prediction tasks. Open problems: broader computational efficiency and learnability of circuits/dynamics.
AI-written summary · about
chapters (20)
- 0:00Start
- 8:38Learning from Quantum Experiments
- 11:44Quantum State Tomography
- 15:18Benchmark Quantum System
- 17:34Palichana Democracy
- 19:19Predicting Ground States
- 24:15Designing a Good Learning Algorithm
- 25:16Shadow Tomography
- 30:39Estimating Fidelity
- 32:33Cross-Platform Verification
- 33:00Distributed Quantum Inner Product Estimation
- 33:35Characterizing Topological Order
- 34:35Why Are We Doing Randomized Experiments
- 35:55Tight Bounds for State Tomography within Coherent Measurements
- 42:55Conventional Experiments
- 48:43Algorithm without Quantum Memory and Algorithm with Quantum Memory
- 50:35The Exponential Separation between Learning with and without Content Memory
- 54:18Proof Strategy
- 58:34Open Questions
- 59:16What Causes of Quantum Dynamics or Quantum Circuits Are Actually Officially Learnable
related talks
01h 11mStatistical Physics and Computation in High Dimension
01h 04mLearning Beyond Stabilizer States
44mQuantum machine learning: Prospects and challenges
01h 31mLightning Talks
59mQuantum Physics for Software Developers
47mQuantum Computing and You
01h 01mQuantum Computing Concepts
01h 15mImproved Machine Learning Algorithm for Predicting Ground State Properties