BEGIN:VCALENDAR
VERSION:2.0
CALSCALE:GREGORIAN
PRODID:UW-Madison-Physics-Events
BEGIN:VEVENT
SEQUENCE:4
UID:UW-Physics-Event-6640
DTSTART:20211103T160000Z
DTEND:20211103T171500Z
DTSTAMP:20260414T193035Z
LAST-MODIFIED:20211012T112820Z
LOCATION:Chamberlin 5280 (Zoom link also available for online particip
 ants who signed up on our mailing list)
SUMMARY:Self-Supervised Learning of Generative Spin-Glasses with Norma
 lizing Flows\, Physics ∩ ML Seminar\, Gavin Hartnett\, RAND
DESCRIPTION:Spin-glasses are universal models that can capture complex
  behavior of many-body systems at the interface of statistical physics
  and computer science\, including discrete optimization\, inference in
  graphical models\, and automated reasoning. In this talk\, I will dis
 cuss the problem of using normalizing flows to build generative models
  of spin-glasses. I will begin with a brief introduction to spin-glass
 es\, and then discuss how the Hubbard-Stratonovich transformation may 
 be used to convert the discrete Boltzmann distribution of a spin-glass
  into a continuous probability density. I will then discuss the proble
 m of modeling the resulting continuous spin-glass using normalizing fl
 ows. Two approaches will be considered\, one based on the forward KL d
 ivergence and one based on the reverse KL divergence. To evaluate both
  approaches\, I will present numerical results for the Sherrington-Kir
 kpatrick spin-glass\, which is known to exhibit rich phenomena such as
  replica symmetry breaking and ultrametricity. The forward KL approach
  is able to approximately capture these phenomena\, whereas the revers
 e KL approach suffers from mode collapse. I will conclude with a discu
 ssion of the physical interpretation of the learned normalizing flow.
URL:https://www.physics.wisc.edu/events/?id=6640
END:VEVENT
END:VCALENDAR