Finance Optimal Execution & Deep RL Project

This project explores optimal execution of portfolio transactions using the Almgren-Chriss model and modern deep reinforcement learning (DRL) methods.

Contents

• Jupyter Notebooks: Step-by-step explanations and experiments on optimal liquidation, trading lists, efficient frontier, and DRL approaches.
• Python Modules: Custom environments and agent implementations for multi-agent and single-agent RL.
• Experiments: Scripts for training and evaluating DDPG, SAC, TD3, PPO agents, and comparison with analytical solutions.
• Results: Scripts and code for plotting and analyzing experiment results.

Main Folders & Files

• finance/
  • *.ipynb — Notebooks for theory, simulation, and DRL experiments.
  • MultiAgent_simulation/ — Multi-agent RL environment and training scripts.
  • syntheticChrissAlmgren.py — Almgren-Chriss simulation environment.
  • ddpg_agent.py, sac_agent.py, td3_agent.py, ppo_agent.py — RL agent implementations.
  • train.py, run_experiments.py, optimize.py — Training and experiment scripts.
  • utils.py — Helper functions for experiments and plotting.
  • Training results for each RL agent can be accessed in the 4-DRL.ipynb notebook.

Project Highlights

• Almgren-Chriss optimal execution model.
• Efficient frontier visualization.
• Deep RL agents (DDPG, SAC, TD3, PPO) for optimal trading.
• Multi-agent market simulation.
• Experiment tracking and result analysis.

License

MIT License

---