news News

Accepted papers at ICAIF `23
Excited to share our recent research on deep hedging and calibration of market data.

We’re excited to announce the acceptance of two papers at The Association for Computing Machinery (ACM) International Conference on AI in Finance (ICAIF) 2023. We would like to thank our partners at NVIDIA and the Hong Kong Exchanges and Clearing Ltd. (HKEX) for their contributions to this research.

Deeper Hedging: A New Agent-based Model for Effective Deep Hedging

We introduce a new agent-based model of market dynamics that combines the traditional Heston model with the extended Chiarella model of the behaviour of market participants. By combining this model with deep reinforcement learning, we demonstrate that our model can outperform the baseline for most transaction cost levels.

Deep Calibration of Market Simulations using Neural Density Estimators and Embedding Networks

By combining deep learning with our market simulators, we demonstrate a novel approach for calibrating models of market dynamics. We use simulation-based inference (SBI) to estimate the posterior of model parameters, using an embedding network to reduce market data into low dimensional summary data, without having to rely on stylised facts or other hand-chosen summary details.

We invite discussions on these topics and more at ICAIF ’23. Reach out if you’d like to delve deeper into these innovations.

Miles Dyson