In the finance industry a common practice is to use something called hedging to reduce risk of an existing position by taking an offsetting position in a different asset. For example, an investor can use a derivative such as a Put or a Call to offset potential losses in the future change of price in the underlying asset. Although classical hedging algorithms can do well in idealized markets, they may not perform as well in real world markets that can incur account transaction costs, liquidity issues, trading restrictions, and other issues.

In recent years, a framework called Deep Hedging has been developed to deal with these issues using an approach called reinforcement learning. This research performed by QC Ware and the Global Technology Applied Research center of JPMorgan Chase & Co. explored using quantum deep learning algorithms to implement a more efficient quantum reinforcement learning technique for Deep Hedging. The potential benefit is to be able to train a quantum neural network with fewer trainable parameters than an equivalent classical approach and improve accuracy and trainability of models that would run on high-performance GPU hardware. The hardware platforms that was used in the research were the Quantinuum H1-1 and H1-2 trapped-ion quantum processors.

Additional information about this research is available in a press release provided by QC Ware that is located here and a technical paper posted on arXiv here.

March 30, 2023