Benchmarking Robustness of Deep Reinforcement Learning approaches to Online Portfolio Management
This work addresses the need for more rigorous evaluation of DRL methods in financial applications, highlighting their limitations for investors and researchers.
The authors tackled the problem of evaluating the robustness of deep reinforcement learning (DRL) algorithms in online portfolio management, finding that most algorithms were not robust, with strategies generalizing poorly and degrading quickly during backtesting.
Deep Reinforcement Learning approaches to Online Portfolio Selection have grown in popularity in recent years. The sensitive nature of training Reinforcement Learning agents implies a need for extensive efforts in market representation, behavior objectives, and training processes, which have often been lacking in previous works. We propose a training and evaluation process to assess the performance of classical DRL algorithms for portfolio management. We found that most Deep Reinforcement Learning algorithms were not robust, with strategies generalizing poorly and degrading quickly during backtesting.