Reinforcement Learning with Parameterized Actions

09/05/2015

∙

We introduce a model-free algorithm for learning in Markov decision processes with parameterized actions-discrete actions with continuous parameters. At each step the agent must select both which action to use and which parameters to use with that action. We introduce the Q-PAMDP algorithm for learning in these domains, show that it converges to a local optimum, and compare it to direct policy search in the goal-scoring and Platform domains.

READ FULL TEXT

Reinforcement Learning with Parameterized Actions

Multi-Pass Q-Networks for Deep Reinforcement Learning with Parameterised Action Spaces

SlateFree: a Model-Free Decomposition for Reinforcement Learning with Slate Actions

Thompson Sampling for Parameterized Markov Decision Processes with Uninformative Actions

Utilizing Skipped Frames in Action Repeats via Pseudo-Actions

Parameterized Reinforcement Learning for Optical System Optimization

Online Reinforcement Learning for Real-Time Exploration in Continuous State and Action Markov Decision Processes

Act-Then-Measure: Reinforcement Learning for Partially Observable Environments with Active Measuring

Reinforcement Learning with Parameterized Actions

Related Research

Multi-Pass Q-Networks for Deep Reinforcement Learning with Parameterised Action Spaces

SlateFree: a Model-Free Decomposition for Reinforcement Learning with Slate Actions

Thompson Sampling for Parameterized Markov Decision Processes with Uninformative Actions

Utilizing Skipped Frames in Action Repeats via Pseudo-Actions

Parameterized Reinforcement Learning for Optical System Optimization

Online Reinforcement Learning for Real-Time Exploration in Continuous State and Action Markov Decision Processes

Act-Then-Measure: Reinforcement Learning for Partially Observable Environments with Active Measuring