Generalized Deep Reinforcement Learning for Trading

This paper proposes generalized deep reinforcement learning with multivariate state space, discrete rewards, and adaptive synchronization for trading any stock held in the S&P 500. Specifically, proposed model observes daily historical data of a 500 multiple market-indicating securities (SPY, IEF, EUR=X, GSG), selects action, reward that is based on correctness selected action independent volatility stocks. The model’s reward-maximizing behavior optimized by using standard q-network (DQN) stabilizes enables to track performance generalizing new experiences from each stock. was trained top 50 holdings tested 100 starting 2006 2022. Experimental results suggest significantly outperforms 100% long-strategy benchmark terms annualized return, Sharpe ratio, maximum drawdown.