AI RESEARCH
A Harmonic Mean Formulation of Average Reward Reinforcement Learning in SMDPs
arXiv CS.AI
•
ArXi:2605.04880v1 Announce Type: cross Recent research has revived and amplified interest in algorithms for undiscounted average reward reinforcement learning in infinite-horizon, non-episodic (continuing) tasks. Semi-Marko decision processes (SMDPs) are of particular interest. In SMDPs, discrete actions stochastically generate both rewards and durations, and the objective is to optimize the average reward rate. Existing algorithms approach this by optimizing the ratio of rewards to durations.