Conjugate Markov Decision Processes

30 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.

Routing Networks and the Challenges of Modular and Compositional Computation (2019, background): “In this case, the routing network is the policy, and can consequently be modeled as a coagent network [Thomas and Barto, 2011].”
A Meta-MDP Approach to Exploration for Lifelong Reinforcement Learning (2019, methodology): “Thomas and Barto [23] proposed the conjugate MDP framework, in which agents solving meta-MDPs (called CoMDPs) can search for the state representation, action representation, or options that maximize the expected return when used by an RL agent solving a single MDP.”
Learning Action Representations for Reinforcement Learning (2019, background): “To learn binary motor primitives, Thomas and Barto [2011] showed how a policy can be modeled as a composition of multiple “coagents”, each of which learns using only the local policy gradient information [Thomas, 2011].”
Policy Gradient Coagent Networks (2011, methodology): “Thomas and Barto [25] call each Ai a coagent and the entire network a coagent network.”
On the Role of Weight Sharing During Deep Option Learning (2019, background): ““Coagent Networks” (CN) were introduced in Thomas & Barto (2011), and model the mapping of any kind of policy to a graph of computation nodes, so-called coagents .”