Reproduce “Dynamic Observation Policies in Observation Cost-Sensitive Reinforcement Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Colin Bellinger, Mark Crowley, Isaac Tamblyn
DMSOA lets an agent choose a control action and how many steps to repeat it before measuring the environment again. The authors compare it with OSMBOA, using Double DQN and prioritized replay across 20 seeds. In Pong, DMSOA’s plotted curves finish higher for wins, costed reward, and intrinsic reward; this does not establish lower elapsed time or computation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.