Rodney Hoskinson’s research on xVA featured at international quantitative finance conferences
Written on 08 November 2019.
Rodney Hoskinson, EDHEC PhD (2016), Director, Quantitative Support (Strategic Trading and Funding), ANZ Banking Group gave an extended talk on “Balance Sheet XVA by Deep Learning and GPU”, joint work* with Stéphane Crepey, University of Evry at The 15th Quantitative Finance Conference in Rome, Italy on 18 October 2019, in the XVA AAD MVA and Initial Margin, GPU stream
Rodney will also be presenting this research work at the first Risk.Net Quant Summit Asia in Singapore on 3 December 2019:
*Abstract: Two competing XVA paradigms are a semi-replication framework and a cost-of-capital, incomplete market approach. Burgard and Kjaer once dismissed an earlier incarnation of the Albanese and Crépey holistic, incomplete market XVA model as being elegant but difficult to solve explicitly. We show that the model (set on a forward/backward SDE formulation) is not only elegant, but also able to be solved efficiently using GPU computing combined with AI methods in a whole bank balance sheet context. We calculate the Mark-to-Market process cube (or its increment, in the context of trade incremental XVA computations) using GPU computing and the XVA process cube using Deep Learning (including joint ES and VaR) Regression methods.