Value-based Q-learning for continuous control can discretise the action space, but the resolution of that discretisation trades early exploration against final precision. Growing Q-Networks (GQN) address this by growing the resolution uniformly over training, according to the arXiv abstract.

The new study, Selective Growing Q-Networks (SGQN), is described as an analysis of adaptive action grids in decoupled Q-learning. Its title suggests a selective approach, in contrast to GQN's uniform growth, though the abstract does not yet specify the exact mechanism or results.

As a single-source summary, this article reflects only the abstract's opening statement. No further details about the method or findings are available from the source provided.