Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

Concordia University

$31,000.00

Risk-aware reinforcement learning for financial portfolio optimization

EF 2024-25

Résumé
Reinforcement learning (RL) consists of a set of techniques solving problems where a sequence of decisions must be optimized so as to achieve a pre-determined goal. RL proceeds through sophisticated trial and error, using experience gained to refine actions the learning agent takes in different contexts. Portfolio optimization is a direct financial application of RL techniques, as it involves periodically rebalancing a portfolio (i.e. adjusting its composition) to obtain optimal statistical properties for its return. Whereas some financial portfolios are designed purely for investment purposes, others are constructed for risk management purposes to offset liabilities of a financial institution or an insurance company. The research program of this proposal places special emphasis on deep hedging methods, which are RL algorithms based on deep neural networks that aim to optimize the synthetic replication of financial positions such as options on stocks, a duty necessary for risk management. Whereas traditional RL algorithms aim at maximizing the average reward an agent receives, financial problems require considering risk, which is represented as the frequency and magnitude of extreme adverse events. This led to the development of so-called risk-aware RL methods which attempt to minimize risk or optimize risk-reward tradeoffs rather than simply maximizing average rewards. The main goal of this proposal is to contribute to the development and improvement of risk-aware RL algorithms for portfolio optimization. The sub-objectives of the grant consist in (i) improving the stability, accuracy and speed of algorithms, (ii) determining the impact of the mathematical formulation of the problem (e.g. the objective function) on performance, (iii) improving the simulation of risk drivers for financial RL algorithms, and (iv) studying the pricing of financial derivatives and options based on such RL algorithms. While several prominent financial institutions have started using RL methods in their operations (for instance J.P. Morgan applies deep hedging), the implementation of these algorithms in the entire industry is still in its infancy. This research program aims to further develop and promote such methods to help financial institutions and insurers improve their operations. Enhancing the stability and safety of the financial and insurance industries also provide benefits to individual citizens, as a more resilient financial system limits the risk of financial crises. Furthermore, limiting the risk exposure faced by life insurers when issuing insurance policies improves the risk-return profile of such policies, thereby allowing insurers to issue them at cheaper rates. As such RL developments for the insurance industry (in particular life insurance) could improve accessibility to insurance for the whole society.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
RGPIN-2024-04593
Name nom
Godin, Frédéric
Department département
Mathematics and Statistics
Organization id
70
Institution établissement
Concordia University
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
31000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1508
Committee name en
Mathematics and Statistics
Committee name fr
Mathématiques et statistique
Area of application code
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
Research subject code
2800
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Artificial intelligence (computer vision, use2603)
Research subject fr
Intelligence artificielle (vision artificielle, utiliser 2603)
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Risk-aware reinforcement learning for financial portfolio optimization
Keywords
quantitative risk management; financial mathematics; reinforcement learning; options and financial derivatives; hedging; portfolio optimization; variable annuities and life insurance; financial time series
Application summary
Reinforcement learning (RL) consists of a set of techniques solving problems where a sequence of decisions must be optimized so as to achieve a pre-determined goal. RL proceeds through sophisticated trial and error, using experience gained to refine actions the learning agent takes in different contexts. Portfolio optimization is a direct financial application of RL techniques, as it involves periodically rebalancing a portfolio (i.e. adjusting its composition) to obtain optimal statistical properties for its return. Whereas some financial portfolios are designed purely for investment purposes, others are constructed for risk management purposes to offset liabilities of a financial institution or an insurance company. The research program of this proposal places special emphasis on deep hedging methods, which are RL algorithms based on deep neural networks that aim to optimize the synthetic replication of financial positions such as options on stocks, a duty necessary for risk management. Whereas traditional RL algorithms aim at maximizing the average reward an agent receives, financial problems require considering risk, which is represented as the frequency and magnitude of extreme adverse events. This led to the development of so-called risk-aware RL methods which attempt to minimize risk or optimize risk-reward tradeoffs rather than simply maximizing average rewards. The main goal of this proposal is to contribute to the development and improvement of risk-aware RL algorithms for portfolio optimization. The sub-objectives of the grant consist in (i) improving the stability, accuracy and speed of algorithms, (ii) determining the impact of the mathematical formulation of the problem (e.g. the objective function) on performance, (iii) improving the simulation of risk drivers for financial RL algorithms, and (iv) studying the pricing of financial derivatives and options based on such RL algorithms. While several prominent financial institutions have started using RL methods in their operations (for instance J.P. Morgan applies deep hedging), the implementation of these algorithms in the entire industry is still in its infancy. This research program aims to further develop and promote such methods to help financial institutions and insurers improve their operations. Enhancing the stability and safety of the financial and insurance industries also provide benefits to individual citizens, as a more resilient financial system limits the risk of financial crises. Furthermore, limiting the risk exposure faced by life insurers when issuing insurance policies improves the risk-return profile of such policies, thereby allowing insurers to issue them at cheaper rates. As such RL developments for the insurance industry (in particular life insurance) could improve accessibility to insurance for the whole society.