Brains, behavior, and robotics

By James S. Albus. 1981

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

Fitted Q-Learning for Relational Domains (2020, methodology): “While the original method (Albus 1981) also used Bellman error to identify the tiles, it was a heuristic based method.”
Adaptive Tile Coding for Value Function Approximation (2007, methodology): “We contribute an adaptive method for tile coding (Albus 1981), a simple, computationally efficient function approximator that has enjoyed broad empirical success (Sutton 1996; Stone, Sutton, &Kuhlmann 2005).”
Algorithms for Reinforcement Learning (2010, background): “Another choice that leads to binary features is tile coding (originally called CMAC, Albus, 1971, 1981).”
Autonomous transfer for reinforcement learning (2008, methodology): “Conceptually, Q(st, at) = Q4DCMAC(st, at) + Q2DCMAC(χX(st), χA(sa)).”
Batch reinforcement learning in a complex domain (2007, background): “CMAC [ 1 ] is a linear function approximation scheme that partitions the input space into axis-aligned regions called tiles.”
Transferring Instances for Model-Based Reinforcement Learning (2008, methodology): “To translate a source task tuple into an appropriate target t ask tuple (Step II) we utilize inter-task mappings [20], which have been 5 Specifically, we used a CMAC [ 1 ] function approximator with 1 4 4-dimensional linear tilings, which is analogous to how Singh and Sutton [15] used 14 2-d dimensi”
Transfer learning for reinforcement learning on a physical robot (2010, methodology): “Specifically, Sutton’s work describes using cerebellar model arithmetic computers (CMACs) [2] as a function approximator for generalizing learning, allowing the agent to generalize across similar states and handle larger (or infinite) state spaces.”
Reinforcement learning with supervision by a stable controller (2004, background): “, CMAC neural network [15], with a total of 25 tilings, or layers, per CMAC.”
Keepaway Soccer: From Machine Learning Testbed to Benchmark (2005, methodology): “CMACs are a form of linear tile-coding function approximation that have been successfully used in many reinforcement learning systems [1].”
Transfer Learning via Inter-Task Mappings for Temporal Difference Learning (2007, methodology): “In one implementation, we use linear tile-coding function approximation, also known as a CMAC (cerebellar model arithmetic computer), which has been successfully used in many reinforcement learning systems (Albus, 1981), including past Keepaway research (Stone et al.”
Scaling Reinforcement Learning toward RoboCup Soccer (2001, methodology): “We used the SMDP version of the Sarsa(λ) algorithm with linear tile-coding function approximation (also known as CMACs) and replacing eligibility traces (see Albus, 1981; Rummery & Niranjan, 1994; Sutton & Barto, 1998).”