Improving Action Selection in MDP's via Knowledge Transfer

By Alexander A. Sherstov, Peter Stone. 2005

94 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Transfer in Reinforcement Learning: A Framework and a Survey (2012, background): “A different approach to representation-transfer involving the action space is proposed in (Sherstov and Stone, 2005).”
Autonomous Agents and Multi-agent Systems Manuscript No. Learning Potential Functions and Their Representations for Multi-task Reinforcement Learning (unknown): “While not explicitly concerned with representation learning, Sherstov and Stone [62] construct a task-independent model of the transition and reward dynamics by identifying shared outcomes and state classes between tasks, and using this for action transfer.”
Using Bisimulation for Policy Transfer in MDPs (2010, background): “Sherstov & Stone (2005) show how policies learned previously can be used to restrict the policy space in MDPs with many actions.”
What can I do here? A Theory of Affordances in Reinforcement Learning (2020, background): “A related family of approaches involves learning when to prune actions (Even-Dar et al., 2003; Sherstov & Stone, 2005; Rosman & Ramamoorthy, 2012; Zahavy et al., 2018) in order to cope with large action-spaces.”
Transfer Learning for Reinforcement Learning Domains: A Survey (2009, background): “For instance, when transferring between maze tasks, basis functions could be learned (Ferguson and Mahadevan, 2006) in the source task, a set of actions to transfer could be selected after training on a set of additional generated source tasks (Sherstov and Stone, 2005), and then parts of different ”
Agent-Agnostic Human-in-the-Loop Reinforcement Learning (2017, background): “Such techniques have been shown to accelerate learning and planning time [39, 19, 38, 2].”
Explainable Sentiment Analysis with Applications in Medicine (2018, background): “The generated explanation, given in terms of state-actions and rewards, are translated into a more descriptive state classes and outcomes by following the approaches of [61] and [62].”
Probabilistic policy reuse in a reinforcement learning agent (2006, background): “The MDP’s formalism is not expressive enough to represent all the concepts involved in knowledge transfer [11], so we define domain and task separately to handle different tasks executed in the same domain.”
Exploring compact reinforcement-learning representations with linear regression (2009, methodology): “This form of generalization has been used to encode many different types of environments in RL, including stochastic STRIPS [12], Object Oriented MDPs (OOMDPs) [5], and typed dynamics [13].”