FI Computational Methods and Data Science Journal Club
Flatiron Institute, 162 5th Avenue
Speaker: Bamdad Hosseini (UW)
Title: Bayesian inference of functions: from theory to computations
Abstract: Bayesian inference is a workhorse of data science but it is often applied in a finite-dimensional/parametric settings. In this talk I will discuss how Bayesian inference can be performed in the non-parametric setting focusing on the inference of functional parameters that belong to infinite-dimensional spaces. I will give a brief review of theory focusing on well-posedness of Bayes’ rule before discussing various algorithmic approaches for computing posterior statistics. I will present function space MCMC algorithms as well as more recent variational techniques inspired by generative models in machine learning.
Attendee Instructions:
FI employees are welcome to attend in person. Please email [email protected] for the Zoom link if you wish to attend remotely.
Visitors (w/out an FI badge) please email [email protected] to be registered to the building or to obtain Zoom information. All visitors must be vaccinated for entry into our buildings.