Michael Gutmann 

University of Edinburgh


Statistical applications of contrastive learning


Statistics Seminar


15th October 2021, 3:00 pm – 4:00 pm
Virtual Seminar, Zoom link: TBA


My talk has two main parts: The first part introduces
contrastive learning as an intuitive and computationally feasible
alternative to likelihood-based learning. The second part looks at
three diverse applications of contrastive learning: parameter
estimation for energy-based models, Bayesian inference for implicit
models, as well as Bayesian experimental design.





Organiser: Henry Reeve

Comments are closed.
css.php