Dynamic Portfolio Optimization with Proximal Policy Optimization

Reinforcement learning approach to daily asset allocation across ETFs with transaction costs, using PPO to learn adaptive allocation policies. (Filler content)

Overview

This project frames portfolio management as an MDP and applies PPO to learn continuous allocation weights across a set of ETFs under transaction costs and realistic constraints. (Filler content)

Methods

• State: market indicators and portfolio features (filler)
• Action: allocation over ETFs + cash (filler)
• Reward: daily return minus transaction costs (filler)
• Algorithm: PPO with clipping and constraints (filler)

Poster

Technologies Used

PPO

Reinforcement Learning

PyTorch

Python

Finance