Reinforcement Learning for Finance Workshop @ ODSC London 2024

Gist bit.ly/odsc_ldn_2024 | DON'T FORGET TO SUBSCRIBE | This workshop at ODSC London in 2024 is based on my new O'Reilly book. Check out the CPF Program for much more content on this topic (growth.tpq.io/cpf).

This video provides a comprehensive introduction to reinforcement learning (RL) in finance, drawing parallels between classic games like chess and real-world financial applications.
● The session kicks off with a basic explanation of dynamic programming, a key concept in finance involving sequential decision-making over time.
● Reinforcement learning, like humans learning through trial and error, allows agents (pieces of software) to interact with financial environments and learn optimal strategies.
● The video uses simple examples like coin flips and dice rolls to illustrate how agents can learn from past experiences and update their strategies based on rewards and penalties.
● The core of the presentation focuses on deep Q-learning, where deep neural networks are used to approximate the agent's policy and handle complex financial scenarios.
● The speaker demonstrates a practical example of training an agent to predict the next day's direction of the Euro-US dollar exchange rate.
● Addressing a common concern in finance, the video tackles the challenge of limited data by introducing Generative Adversarial Networks (GANs).
● GANs are used to create synthetic financial time series data that is statistically indistinguishable from the real data, thus expanding the training data for the RL agent.
● The presentation concludes with a demonstration of how GANs can generate synthetic returns for a gold ETF (GLD), and statistical tests are used to confirm the similarity between the synthetic and real data.
The video provides valuable insights into how reinforcement learning and GANs can be applied to solve practical financial problems, offering viewers a foundational understanding of these cutting-edge technologies.