Information-Theoretic Generalization Bounds for Quantum Learning

Watch a conference presentation from TQC 2024 exploring a unified mathematical framework for quantum learning theory and generalization bounds. Delve into how quantum learning tasks - from state discrimination to quantum PAC learning - can be analyzed through an information-theoretic lens. Learn about new bounds on expected generalization error expressed through classical and quantum mutual information quantities, established using quantum optimal transport and concentration inequalities. Discover how this framework provides intuitive generalization bounds across various quantum learning scenarios, creating a foundation for understanding quantum learning through quantum information theory. Presented by researchers Matthias C. Caro, Tom Gur, Cambyse Rouzé, Daniel Stilck França, and Sathyawageeswar Subramanian at the 19th Conference on Theory of Quantum Computation, Communication and Cryptography held at OIST, Japan.