Split Feature Space Ensemble Method using Deep Reinforcement Learning for Algorithmic Trading

Ensemble Reinforcement Learning: A Survey

“…In this framework, the base learner completes the final prediction step by step in a certain order [65]. [14] proximal policy optimization, advantage actor-critic, deep deterministic policy gradient 2020 Saphal et al [68] advantage actor-critic, sample efficient actor-critic with experience replay, actor-critic using Kronecker-factored trust region, deep deterministic policy gradient, soft actor-critic, trust region policy optimization 2021 Smit et al [28] double deep Q-Learning, soft actor-critic 2022 Eriksson et al [69] residual gradient, TD, TD(λ) 2022 Németh, Marcell and Szűcs, Gábor [70] deep deterministic policy gradient, advantage actorcritic, proximal policy optimization new method of combining online and offline training algorithms or using training algorithms based on different optimization strategies, which can take advantage of their respective strengths to handle complex tasks. Accordingly, the complexity of ERL methods using such improved strategies increases.…”

“…This system architecture allows the ensemble models in ERL to handle the same or different tasks. [95] fuel economy improvement Q-learning 2021 Carta et al [49] stock market forecasting deep Q-Network 2022 Li et al [58] regional GDP prediction deep Q-Network 2022 Zijie Cao and Hui Liu [59] carbon price forecasting Q-learning 2022 Németh, Marcell and Szűcs, Gábor [70] algorithmic trading proximal policy optimization, advantage actorcritic, deep deterministic policy gradient…”