Reinforcement Learning Methods and Applications in Finance
Summary
This Chinese-language summary presents a 2021 survey of reinforcement learning in financial decision-making. It introduces Markov decision processes as a foundation, then identifies value-based and policy-based learning approaches, including extensions that use neural networks for deep reinforcement learning. The survey is positioned against classical stochastic control: simpler models can be tractable but may miss market behavior, while richer models can become difficult to solve with traditional tools.
The article maps applications across optimal execution, portfolio optimization, option pricing and hedging, market making, smart order routing, and robo-advisory. It describes the field’s motivation as learning from abundant trading and order-flow data while relying less heavily on specified market models. The source is a literature survey, not a single trading strategy or a controlled performance comparison. The excerpt supplies no quantitative results, and its overview does not establish that reinforcement learning will outperform classical methods in any particular market or deployment setting.
Key ideas
- Reinforcement learning frames financial decisions as actions learned through interaction with a market or simulated environment.
- Markov decision processes underpin many of the methods reviewed in the survey.
- The survey covers value-based, policy-based, and deep reinforcement learning approaches.
- Applications include execution, portfolio allocation, derivatives, market making, order routing, and robo-advisory.
- The review describes reduced reliance on explicit models as a motivation, but does not claim universal superiority over stochastic control.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.