Feb, 2024
分层变压器是高效的元强化学习者
Hierarchical Transformers are Efficient Meta-Reinforcement Learners
TL;DRHierarchical Transformers for Meta-Reinforcement Learning (HTrMRL) is an online meta-reinforcement learning approach that improves learning efficiency and generalization capabilities, outperforming the previous state-of-the-art algorithm in various tasks.