BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Simplicity bias in random design - Ard Louis (University of Cambri
 dge\; University of Oxford)
DTSTART:20190513T131000Z
DTEND:20190513T135000Z
UID:TALK124279@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:The design of a soft-matter system can be recast as an input-o
 utput map\, where the inputs are the parameters that fix the components an
 d their interactions\, and the outputs describe the outcome of a self-asse
 mbly process. By extending the coding theory from algorithmic information 
 theory\, we have recently shown [K Dingle\, C. Camargo and AAL\, Nat Comm.
  9\, 761 (2018)] that for many computable maps\, the a priori probability 
 P(x) that randomly sampled inputs generate a particular output x decays ex
 ponentially with the approximate Kolmogorov complexity $\\tilde{K}(x)$&nbs
 p\;of that output. While Kolmogorov complexity is technically uncomputable
 \, we show how to make approximations that work in practice\, allowing for
  a tight upper bound on P(x). For soft matter systems\, simplicity bias im
 plies that randomly sampling design inputs will naturally lead to outputs 
 that have low descriptional complexity. Since high symmetry structures typ
 ically have low descriptional complexity\, simplicity bias implies that ra
 ndomly picking design patterns can lead to the spontaneous emergence of hi
 ghly symmetric self-assembled structure. We provide evidence for these tre
 nds for self-assembled RNA and protein structures.&nbsp\;
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR