BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A Data-Driven Market Simulator for Small Data Environments - Blank
 a  Horvath  (King's College London\; JP Morgan)\; Magnus  Wiese (JP Morgan
 )
DTSTART:20210316T151000Z
DTEND:20210316T153500Z
UID:TALK157945@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:<span>Applying deep reinforcement learning (DRL) algorithms su
 ch as Deep Hedging to financial markets relies on the availability of larg
 e amounts of realistic market data. On a daily time scale the amount of fi
 nancial data for a single underlying is insufficient for training a DRL ag
 ent. In this talk\, two novel approaches to simulating financial markets b
 y using a rough paths perspective are presented. The first market simulato
 r presented by Dr. Blanka Horvath pairs conditional variational autoencode
 rs (cVAEs) with signatures and allows sampling realistic spot price paths 
 conditional on the market&rsquo\;s state by inverting sampled market signa
 tures. Afterwards\, Magnus Wiese presents a market simulator for generatin
 g the dynamics of spot price paths and a high-dimensional grid of discrete
  local volatilities (DLVs) in a robust fashion. Here\, the market simulato
 r is split into two learning modules: a compression algorithm for encoding
  the grid of DLVs into a low-dimensional orthonormal representation by lev
 eraging autoencoders and signature cumulants\, and a generative modelling 
 algorithm for learning to generate spot prices and the encoded grid of DLV
 s conditional on market information.<b></b></span>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR