Reproduce “Adaptive and Multiple Time-scale Eligibility Traces for Online Deep 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
Taisuke Kobayashi
The authors propose generalized eligibility traces for online deep actor–critic learning, adapting trace decay to changes in policy and value outputs and combining short- and long-timescale traces. In four stationary PyBullet tasks, the configured combined method outperformed tested alternatives; adaptive decay alone showed no clear benchmark gain. A real-robot case study showed learning after a reward-target swap, not changed dynamics.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.