BEGIN:VCALENDAR
PRODID:-//Columba Systems Ltd//NONSGML CPNG/SpringViewer/ICal Output/3.3-
 M3//EN
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
DTSTAMP:20230207T223228Z
DTSTART:20230214T150000Z
DTEND:20230214T160000Z
SUMMARY:Optimizing LES closure models through Reinforcement Learning
UID:{http://www.columbasystems.com/customers/uom/gpp/eventid/}dz2-ldut0v2
 e-lieie7
DESCRIPTION:Please note that this is an online-only event.\n\nAbstract of
  lecture:\n\nReinforcement learning (RL) is considered as the third lear
 ning paradigm\, besides unsupervised and supervised learning.  In RL\, t
 he learning task is framed as a Markov Decision Process (MDP)\, which is
  solved by an optimal policy. This policy is either approximated directl
 y or through the evaluation of a learned value action function. The lear
 ned policy represents the current control strategy for solving the MDP. 
 Its parameters are updated through repeated sampling of the policy’s pro
 posed action space through interaction with the environment of the MDP\,
  which emits reward signals intermittently and estimating the gradient o
 f the objective w.r.t these parameters. This optimization within the con
 text of a dynamical system makes the RL approach somewhat orthogonal to 
 supervised learning (SL) in that no training samples need to be known a 
 priori\, only a definition of a meaningful reward (which could be a sing
 le scalar value) is necessary. This more indirect guidance of the learni
 ng process makes RL methods relatively sample-inefficient and training s
 tability is less well understood than for SL methods\, however\, its pos
 sible benefits have been demonstrated in a range of applications from au
 tonomous driving\, strategic games and flow control. \nIn this talk\, I 
 will present data-driven approaches to LES modeling for implicitly filte
 red high order discretizations. Wheres supervised learning of the Reynol
 ds force tensor based on non- ocal data can provide highly accurate resu
 lts that provide higher a priori correlation than any existing closures\
 , a posteriori stability remains an issue. I will give reasons for this 
 and introduce reinforcement learning (RL) as an alternative optimization
  approach. Our initial experiments with this method suggest that is it m
 uch better suited to account for the uncertainties introduced by the num
 erical scheme and its induced filter form on the modeling task. For this
  coupled RL-DG framework\, I will present discretization-aware model app
 roaches for the LES equations (c.f. Fig. 1) and discuss the future poten
 tial of these solver-in-the-loop optimizations. \n\nBio of speaker:\n\nA
 ndrea obtained a M.Sc. degree in aerospace engineering with a focus on f
 luid dynamics from the Georgia Institute of Technology in Atlanta (USA) 
 and a doctoral degree from the University of Stuttgart (Germany) in comp
 utational fluid dynamics (CFD). She held the Dorothea-Erxleben professor
 ship at the Institute of Fluid Dynamics and Thermodynamics of the Otto-v
 on-Guericke University in Magdeburg (Germany) from 2020 to 2022 and is c
 urrently a professor for numerical methods in fluid dynamics at the facu
 lty of aerospace engineering and geodesy at the University of Stuttgart.
  Her areas of interest include numerical discretization schemes for mult
 iscale-multiphysics problems\, in particular high order methods\, high p
 erformance computing and visualization\, Large Eddy Simulation methods a
 nd models\, shock capturing schemes\, uncertainty quantification methods
  and machine learning. She is a co-developer of the open-source high ord
 er Discontinuous Galerkin CFD framework FLEXI. Recent fields of applicat
 ion include uncertainty quantification of feedback loops in acoustics\, 
 particle-laden flow in turbomachines\, wake-boundary layer interaction f
 or transport aircraft at realistic flight conditions\, shock-droplet int
 eractions and data-driven models for LES closures.\n\n
STATUS:TENTATIVE
TRANSP:TRANSPARENT
CLASS:PUBLIC
END:VEVENT
END:VCALENDAR