Michael Gutmann

Wednesday 8th November 2017

Time: 4.00pm

Ground Floor Seminar Room

25 Howland Street, London, W1T 4JG

Efficient Likelihood-Free Inference

We consider the problem of performing statistical inference when
evaluating the likelihood function is prohibitively costly but sampling
from the model is possible. Such inference problems occur in a wide
range of disciplines ranging from infectious disease epidemiology to
econometrics and computer vision. I give an overview of my work and then
explain how we can accelerate the inference substantially by combining
probabilistic modelling with decision making under uncertainty.

Reference paper:
http://jmlr.org/papers/v17/15-017.html