Variable Resolution Dynamic Programming

By Andrew W. Moore. 1991

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

Programmatically Interpretable Reinforcement Learning (2018, methodology): “This implementation is based on the system presented in (Moore, 1991).”
Reinforcement Learning: A Survey (1996, methodology): “Variable Resolution Dynamic Programming The VRDP algorithm (Moore, 1991) enables conventional dynamic programming to be performed in real-valued multivariate state-spaces where straightforward discretization would fall prey to the curse of dimensionality.”
Efficient Learning and Planning Within the Dyna Framework (1993, background): “…combination of dynamic-programming-based planning strategies with learned worldmodels represents a promising approach to e ective learning control (Moore, 1991), andthe computational expense of more conventional planning makes incremental approacheslike Dyna (Sutton, 1990; 1991) especially…”
Environment-Independent Task Specifications via GLTL (2017, background): “For example: • Goal-based tasks like mountain car (Moore, 1991): If p represents the attribute of being at the goal (the top of the hill, say), ♦ p corresponds to eventually reaching the goal.”
Stable Function Approximation in Dynamic Programming (1995, methodology): “…mimic the location of this discontinu-ity exactly will have no representable functions (andtherefore no xed points) within d2 of V .6 EXPERIMENTS: HILL-CAR THEHARD WAYIn the hill-car world (Moore, 1991, Boyan and Moore,1995, Gordon, 1995), the agent must drive a car upto the top of a steep hill.”
A Study of Reinforcement Learning in the Continuous Case by the Means of Viscosity Solutions (2000, unknown): “…(Now e, 1995), (Glorennec & Jou e, 1997)), approximators based on state aggregation (see (Singh, Jaakkola, & Jordan, 1994)), clustering (see (Mahadevan & Connell, 1992)), sparse-coarse-coded functions (see (Sutton, 1996)) and variable resolution grids (see (Moore, 1991), (Moore & Atkeson, 1995)).”
Locally Weighted Learning for Control (1997, methodology): “This can be done heuristically with the local density of the data providing an uncertainty estimate (Moore 1991a) or by making sensible statistical assumptions (Schaal and Atkeson 1994b; Cohn et al. 1995).”
Self-improving reactive agents based on reinforcement learning, planning and teaching (1992, methodology): “…Grefenstette et al. (1990) a genetic algorithm method, Mahadevan and Connell (1991) a statistical clustering method, Chapman and Kaelbling (1991) a method similar to decision-tree algorithms, and Moore (1991) a method which is basically look-up tables but which uses variable state resolution.”
Tree Based Discretization for Continuous State Space Reinforcement Learning (1998, methodology): “Techniques using a non-uniform discretization are referred to as variable resolution techniques (Moore 1991).”