A paper from the Department of Computing has received the Best Student Paper Award at the 6th ACM International Conference on AI in Finance (ICAIF '25) in Singapore.

A paper from the Department of Computing has received the Best Student Paper Award at the 6th ACM International Conference on AI in Finance (ICAIF '25), held in Singapore. The award-winning study titled "ProtoHedge: Interpretable Hedging with Market Prototypes" was led by Dr Ce Guo and advised by Professor Wayne Luk, and originated from Lisa Faloughi's (MSc) individual project within the Department.

The research contributes to the EPSRC SONNETS programme, which aims to develop systems for real-time, national-level risk analysis. The paper addresses a key challenge in deep hedging where reliance on opaque neural networks limits trust and auditability.

We demonstrated that by using market prototypes, we can build hedging agents that are fully interpretable yet achieve financial utility comparable to opaque 'black-box' models. Dr Ce Guo Research Fellow (Department of Computing)

To tackle this issue, the research introduces ProtoHedge. This architecture functions as a deep learning model to retain high modelling capacity but ensures transparency by learning a finite set of representative market states known as prototypes. The system makes hedging decisions via a similarity-based mechanism where the action is a weighted average of learned actions associated with each prototype. This approach allows decisions to be traced back to understandable market scenarios.

The experimental results demonstrate that the model maintains good hedging performance. Specifically, the hedging effectiveness is comparable to that of the original black-box deep hedging with a difference of less than 0.40%. These findings suggest that AI-based financial systems may not need to sacrifice accuracy in achieving interpretability.

Commenting on the achievement, Dr Ce Guo said: "A key finding of our research is that transparency and hedging effectiveness are not mutually exclusive. We demonstrated that by using market prototypes, we can build hedging agents that are fully interpretable yet achieve financial utility comparable to opaque 'black-box' models. We hope this challenges the assumption that opacity is the price we must pay for high-quality risk management."

About ICAIF

ICAIF is a peer reviewed venue supported by the Association for Computing Machinery that brings together researchers to discuss the impact of artificial intelligence on the financial sector.