English

Double Q($\sigma$) and Q($\sigma, \lambda$): Unifying Reinforcement Learning Control Algorithms

Artificial Intelligence 2017-11-07 v1 Machine Learning Machine Learning

Abstract

Temporal-difference (TD) learning is an important field in reinforcement learning. Sarsa and Q-Learning are among the most used TD algorithms. The Q(σ\sigma) algorithm (Sutton and Barto (2017)) unifies both. This paper extends the Q(σ\sigma) algorithm to an online multi-step algorithm Q(σ,λ\sigma, \lambda) using eligibility traces and introduces Double Q(σ\sigma) as the extension of Q(σ\sigma) to double learning. Experiments suggest that the new Q(σ,λ\sigma, \lambda) algorithm can outperform the classical TD control methods Sarsa(λ\lambda), Q(λ\lambda) and Q(σ\sigma).

Keywords

Cite

@article{arxiv.1711.01569,
  title  = {Double Q($\sigma$) and Q($\sigma, \lambda$): Unifying Reinforcement Learning Control Algorithms},
  author = {Markus Dumke},
  journal= {arXiv preprint arXiv:1711.01569},
  year   = {2017}
}
R2 v1 2026-06-22T22:36:22.255Z