Better Hedging Through Deep Learning

Recent work by JPMorgan's Quantitative Research team has developed a new technique for calculating optimal hedges for arbitrary derivatives portfolios that is based on training a deep learning neural network. This technique reduces to standard risk neutral hedges when the standard risk neutral assumptions hold: no transaction costs, no restrictions on hedge size, continuous rebalancing of hedges, and no unhedgeable risks. However, it cleanly generalizes to the cases where those assumptions are violated. This technique shows promise for hedging problems in less liquid markets, such as physical commodity asset optimization and variable annuity hedging.

SCHEDULE:
6:00-6:30pm Networking
6:30-7:30pm Talk
7:30-8:00pm Networking