BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Hybrid Summary Statistics: Telling Neural Networks where to look -
  Lucas Makinen (Imperial)
DTSTART:20241014T150000Z
DTEND:20241014T160000Z
UID:TALK222757@talks.cam.ac.uk
CONTACT:65128
DESCRIPTION:Neural networks can capture an impressive range of patterns in
  supervised and unsupervised settings. This makes them useful for capturin
 g data features from simulations or training datasets that can be used for
  highly informative and in some cases optimal simulation-based inference. 
 However\, asymptotic optimality often requires both enormous networks and 
 large vats of training data. \n\nI will present a way to capture high-info
 rmation posteriors from training sets that are sparsely sampled over the p
 arameter space with smaller networks for robust simulation-based inference
 . In physical inference problems\, we can often apply domain knowledge to 
 define traditional summary statistics to capture some of the information i
 n a dataset. I will show that augmenting these statistics with neural netw
 ork outputs to maximise the mutual information improves information extrac
 tion compared to neural summaries alone or their concatenation to existing
  summaries and makes inference robust in settings with low training data. 
 I will also informally discuss extensions to this information-theoretic fr
 amework that go beyond cosmological parameter inference.
LOCATION:Martin Ryle Seminar Room\, KICC
END:VEVENT
END:VCALENDAR