TradeR: Practical Deep Hierarchical Reinforcement Learning for Trade Execution

Advances in Reinforcement Learning (RL) span a wide variety of applications\nwhich motivate development in this area. While application tasks serve as\nsuitable benchmarks for real world problems, RL is seldomly used in practical\nscenarios consisting of abrupt dynamics. This allows one to rethink the problem\nsetup in light of practical challenges. We present Trade Execution using\nReinforcement Learning (TradeR) which aims to address two such practical\nchallenges of catastrophy and surprise minimization by formulating trading as a\nreal-world hierarchical RL problem. Through this lens, TradeR makes use of\nhierarchical RL to execute trade bids on high frequency real market experiences\ncomprising of abrupt price variations during the 2019 fiscal year COVID19 stock\nmarket crash. The framework utilizes an energy-based scheme in conjunction with\nsurprise value function for estimating and minimizing surprise. In a\nlarge-scale study of 35 stock symbols from the S&P500 index, TradeR\ndemonstrates robustness to abrupt price changes and catastrophic losses while\nmaintaining profitable outcomes. We hope that our work serves as a motivating\nexample for application of RL to practical problems.\n