Learning Representations for Control with Hierarchical Forward Models. (arXiv:2206.11396v1 [cs.LG])

Web: http://arxiv.org/abs/2206.11396

Learning control from pixels is difficult for reinforcement learning (RL)
agents because representation learning and policy learning are intertwined.
Previous approaches remedy this issue with auxiliary representation learning
tasks, but they either do not consider the temporal aspect of the problem or
only consider single-step transitions. Instead, we propose Hierarchical
$k$-Step Latent (HKSL), an auxiliary task that learns representations via a
hierarchy of forward models that operate at varying magnitudes of step skipping
while also learning to communicate between levels …

arxiv hierarchical learning lg models