{"success":1,"msg":"","color":"rgb(28, 35, 49)","title":"Approximate Bayesian computation via variational approximation<\/b>","description":"webinar","title2":"","start":"2021-05-13 17:00","end":"2021-05-13 18:00","responsable":"Giacomo Zanella <\/i><\/a>","speaker":"Yun Yang (Illinois Urbana-Champaign)","id":"43","type":"webinar","timezone":"Europe\/Rome","activity":"Join Zoom Meeting\r\nhttps:\/\/zoom.us\/j\/92037471575?pwd=Y3ZuWkRZQ1BuSitTVE5HMGsxYmdJdz09\r\nMeeting ID: 920 3747 1575\r\nPasscode: 389650","abstract":"Variational inference is a popular alternative to Markov Chain Monte Carlo (MCMC) for approximating complicated probability densities arising from Bayesian hierarchical models. However, unlike MCMC that is guaranteed to produce precise samples from the target density for ergodic chains, theoretical aspects of VI is less explored in the literature. In this talk, we will discuss our recent progress towards theoretical understanding of VI. Firstly, we provide general conditions under which VI is consistent for point estimation and model selection. Secondly, we study the distributional convergence of VI for parameter models. A multiplier bootstrap method is proposed for valid uncertainty quantification. Lastly, we turn to the computational aspects of VI by studying the algorithmic convergence of coordinate descent through the lens of gradient flow in the space of probability measures."}