Speaker: Yingru Li

Title: The Randomized Elliptical Potential Lemma with an Application to Linear Thompson Sampling

Time: Apr 16 2pm-5pm

Short Abstract:

This work generalize the elliptical potential bound to any arbitrary prior and noise distributions.

The second contribution of this note is to apply the aforementioned randomized elliptical potential lemma in combination with the proof techniques in (Dong and Van Roy, 2018; Kalkanli and Özgür, 2020) to prove an \(\mathcal{O}(d \sqrt{T \log T})\) bound for the Bayesian regret of the well-known linear Thompson sampling (LinTS).

Reference:

Hamidi, Nima, and Mohsen Bayati. ”The Randomized Elliptical Potential Lemma with an Application to Linear Thompson Sampling.” arXiv preprint arXiv:2102.07987 (2021).

Kalkanlı, Cem, and Ayfer Özgür. ”An Improved Regret Bound for Thompson Sampling in the Gaussian Linear Bandit Setting.” 2020 IEEE International Symposium on Information Theory (ISIT). IEEE, 2020.