Reinforcement Learning in Continuous Action Spaces

By Hado van Hasselt, Marco A. Wiering. 2007

249 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Control and Systems Engineering

How researchers see this paper

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

In Artificial Intelligence

In Computational Theory and Mathematics

In Control and Systems Engineering

Citations in the corpus

Critic Regularized Regression (2020, methodology): “An earlier version of a similar algorithm can be found in van Hasselt and Wiering [36].”
Survey on reinforcement learning for language processing (2021, methodology): “Such continuous language representations could seamlessly play together with continuous RL algorithms like CACLA [119], Deterministic Policy Gradient (DPG) [105] or deep DPG (DDPG) [73].”
Safe Exploration of State and Action Spaces in Reinforcement Learning (2012, methodology): “It has been shown empirically that this procedure can result in better policies than when step size depends on the size of the TD error (Van Hasselt & Wiering, 2007).”
Exploring parameter space in reinforcement learning (2010, background): “Instead, states are mapped to actions directly by means of a parameterized function approximator, without utilizing Q-values.”
The problem with DDPG: understanding failures in deterministic environments with sparse rewards (2019, methodology): “The prototypical actor-critic algorithm using a model of V as a critic is CACLA (Van Hasselt & Wiering, 2007).”
Using continuous action spaces to solve discrete problems (2009, methodology): “For the mountain car problem [7] we used the same problem description as in the book by Sutton and Barto [10].”
A predictive network architecture for a robust and smooth robot docking behavior (2012, background): “This object feature selection, on the other hand, also provides a cue in the early visual cortex to predict upcoming sensory data via recurrent and topdown connections based on the prior noisy sensory information [3, 4].”
Evolutionary Computation for Reinforcement Learning (2012, methodology): “Of course, actor-critic methods (Doya, 2000; Peters and Schaal, 2008) and other techniques (Gaskett et al, 1999; Millán et al, 2002; van Hasselt and Wiering, 2007) can also be used to make temporal-difference methods suitable for continuous action spaces.”
Reinforcement Learning in Continuous State and Action Spaces (2012, methodology): “In this section we discuss the continuous actor-critic learning-automaton (Cacla) algorithm (van Hasselt and Wiering, 2007, 2009).”
Learning and Using Models (2012, methodology): “Alternatively, one can represent the policy with a function approximator (for example, in an actor-critic method) and update the function approximator appropriately to output the best continuous action (Sutton et al, 1999; van Hasselt and Wiering , 2007).”