講者：李彥寰助理教授（台大資工系助理教授／統計所合聘教師）
時間：111年9月15日（星期四）13:30
地點：台大次震宇宙館601室
講題：Multiplicative gradient algorithms for quantum state estimation
摘要：Quantum state estimation is an important task in quantum information. Interestingly, its maximum-likelihood formulation is exactly the quantum counterpart of the Kelly criterion, an asymptotically optimal long-term investment strategy. In this talk, I will present batch and online algorithms for maximum-likelihood quantum state estimation. The batch algorithm generalizes Cover's 1984 algorithm for the Kelly criterion to the quantum setup; our non-asymptotic analysis complements Cover's asymptotic convergence proof. The online algorithm generalizes Soft-Bayes, an online portfolio selection algorithm Orseau et al. proposed in ALT 2019, to the quantum setup. Curiously, both algorithms share a multiplicative gradient formulation, but their analyses are very different.