Sitting in the Kavli Institute for Cosmology, I am embedded in one of the leading institutes in the world for many aspects of studying the largest scales in physics.
Likelihood free inference, or Simulation based inference (image from summary article by Cranmer et al.) is an emerging sub-field and comes in many flavours,
At it’s core we seek to use modern machine learning to accelerate inference with expensive classical forward models. Many situations in cosmology are an attractive prospect for this sort of work.
Current interests in this field
- How do we get the well calibrated uncertainty we need to rely on ML trained emulators for precise inference? This has strong links to the Bayesian Machine Learning topics discussed here
If you are interested in discussing any area of LFI please get in contact!