Many problems in hig h-dimensional statistics rely on low-rank decompos itions of matrices. Examples include matrix compl etion\, recommender systems or collaborative filt ering\, and graph clustering or community detectio n. Most commonly\, estimates are obtained by solv ing an optimisation problem through SDP relaxatio ns\, expectation maximisation\, or projected gradi ent descent algorithms. Bayesian analogs of these procedures provide estimates of uncertainty\, but these are rarely exploited in practice. In this talk\, we explore how the posterior distribution in matrix factorisation models can be put to use i n sequential design problems. Bayesian procedures such as Thompson sampling and the Bayesian UCB h ave been shown to achieve optimal regret in Multi- Arm Bandit problems. We present a simulation stud y supporting similar strategies in recommender sys tems. LOCATION:Seminar Room 1\, Newton Institute CONTACT:INI IT END:VEVENT END:VCALENDAR