PPO

Deep Reinforcement Learning Strategies in Finance: Insights into Asset Holding, Trading Behavior, and Purchase Diversity ArXiv ID: 2407.09557 “View on arXiv” Authors: Unknown Abstract Recent deep reinforcement learning (DRL) methods in finance show promising outcomes. However, there is limited research examining the behavior of these DRL algorithms. This paper aims to investigate their tendencies towards holding or trading financial assets as well as purchase diversity. By analyzing their trading behaviors, we provide insights into the decision-making processes of DRL models in finance applications. Our findings reveal that each DRL algorithm exhibits unique trading patterns and strategies, with A2C emerging as the top performer in terms of cumulative rewards. While PPO and SAC engage in significant trades with a limited number of stocks, DDPG and TD3 adopt a more balanced approach. Furthermore, SAC and PPO tend to hold positions for shorter durations, whereas DDPG, A2C, and TD3 display a propensity to remain stationary for extended periods. ...

Evaluation of Deep Reinforcement Learning Algorithms for Portfolio Optimisation ArXiv ID: 2307.07694 “View on arXiv” Authors: Unknown Abstract We evaluate benchmark deep reinforcement learning algorithms on the task of portfolio optimisation using simulated data. The simulator to generate the data is based on correlated geometric Brownian motion with the Bertsimas-Lo market impact model. Using the Kelly criterion (log utility) as the objective, we can analytically derive the optimal policy without market impact as an upper bound to measure performance when including market impact. We find that the off-policy algorithms DDPG, TD3 and SAC are unable to learn the right $Q$-function due to the noisy rewards and therefore perform poorly. The on-policy algorithms PPO and A2C, with the use of generalised advantage estimation, are able to deal with the noise and derive a close to optimal policy. The clipping variant of PPO was found to be important in preventing the policy from deviating from the optimal once converged. In a more challenging environment where we have regime changes in the GBM parameters, we find that PPO, combined with a hidden Markov model to learn and predict the regime context, is able to learn different policies adapted to each regime. Overall, we find that the sample complexity of these algorithms is too high for applications using real data, requiring more than 2m steps to learn a good policy in the simplest setting, which is equivalent to almost 8,000 years of daily prices. ...