BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Playing Newton: Learning equations of motion from data - Ilya Neme
 nman (Emory University)
DTSTART:20171016T120000Z
DTEND:20171016T130000Z
UID:TALK77062@talks.cam.ac.uk
CONTACT:Julius Bier Kirkegaard
DESCRIPTION:Arguably\, science' goal of understanding nature can be formul
 ated as inferring mathematical laws that govern natural systems from exper
 imental data. With the fast growth of power of modern computers and of art
 ificial intelligence algorithms\, there has been a recent surge in attempt
 s to automate this goal and to design\, to some extent\, an “artificial 
 scientist.” I will discuss this emerging field\, but will focus primaril
 y on our own approach to it. I will introduce an algorithm that we have re
 cently developed\, which allows one to infer the underlying dynamical equa
 tions behind a noisy time series\, even if the dynamics are nonlinear\, an
 d only a few of the relevant variables are measured. I will illustrate the
  method on applications to toy problems\, including inferring the iconic N
 ewton’s law of universal gravitation\, and dynamics of a few synthetic b
 iochemical systems. I will end with applications to real biological data: 
 modeling calcium dynamics in pancreatic beta cells\, as well as modeling t
 he landscape of possible behavioral states underlying reflexive escape fro
 m pain in a roundworm.  
LOCATION:MR11\, Centre for Mathematical Sciences\, Wilberforce Road\, Camb
 ridge
END:VEVENT
END:VCALENDAR