Barycentric Interpolators for Continuous Space and Time Reinforcement Learning

By R´emi Munos, Andrew W. Moore. 1998

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

Increasing the Action Gap: New Operators for Reinforcement Learning (2015, background): “Thus the consistent Bellman operator remains relevant in cases where P is a deterministic transition kernel, for example when applying multilinear or barycentric interpolation to continuous space MDPs (e.g. Munos and Moore 1998).”
Instabilities of Offline RL with Pre-Trained Neural Representation (2021, background): “See [Gordon, 1995, 1996, Munos and Moore, 1999, Ormoneit and Sen, 2002, Kakade, 2003, Zanette et al., 2019] for discussion on this topic.”
Variable Resolution Discretization in Optimal Control (2002, methodology): “We implement a class of functions known as barycentric interpolators (Munos & Moore 1998), built from a triangulation of the state-space using a tree structure.”
A Study of Reinforcement Learning in the Continuous Case by the Means of Viscosity Solutions (2000, background): “It is easy to prove (see (Munos & Moore, 1998) or (Munos, 1997a)) that assuming some smoothness assumptions (r;R Lipschitzian), the approximated V n satisfy the condition (34) and theorem 6 applies.”
Planning algorithms (2006, methodology): “If barycentric subdivision is used to decompose the cube using the midpoints of all faces, then the point-location problem can be solved in O(n lg n) time [265, 610, 724], which is an improvement over the O(2) scheme described above.”
Cognitive developmental robotics as a new paradigm for the design of humanoid robots (2001, unknown): “[16, 17].”
What are the Statistical Limits of Offline RL with Linear Function Approximation? (2020, background): “l., 2020]. Error Amplication In RL. Error amplication induced by distribution shift and long planning horizon is a known issue in the theoretical analysis of RL algorithms. See [Gordon, 1995, 1996, Munos and Moore, 1999, Ormoneit and Sen, 2002, Kakade, 2003, Zanette et al., 2019] for papers on this ”
Fast deep reinforcement learning using online adjustments from the past (2018, methodology): “Using previous experience at decision time is closely related to non-parametric approaches for Q - function approximation [Santamaría et al., 1997, Munos and Moore, 1998, Gabel and Riedmiller, 2005].”