I presented at the International Fintech Research Conference on January 31, 2025, discussing Optimal Execution via Reinforcement Learning in Agent-Based Simulations.

My presentation focused on optimal execution strategies using reinforcement learning. I introduced the problem definition and its formulation as a Markov Decision Process (MDP), highlighting key aspects such as execution trajectories, market impact, and transaction cost minimization. I then delved into reinforcement learning algorithms, discussing Q-learning and Deep Q-Networks (DQN). Additionally, I explored the role of limit order book (LOB) simulations, comparing various execution policies against traditional approaches like TWAP and passive algorithms. Finally, I presented experimental results showcasing how reinforcement learning can minimize market impact while optimizing execution efficiency.

Download the slides [PDF].

Here you can find the arxiv paper: Optimal Execution via Reinforcement Learning in Agent Based Simulations.