Using locally weighted regression for robot learning

By Christopher G. Atkeson. 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

Increasing the Action Gap: New Operators for Reinforcement Learning (2015, background): “One may ask whether it is possible to extend the consistent Bellman operator to Q-value approximation schemes which lack a probabilistic interpretation, such as linear approximation (Sutton 1996), locally weighted regression (Atkeson 1991), neural networks (Tesauro 1995), or even informationtheoreti”
Autonomous helicopter control using reinforcement learning policy search methods (2001, background): “[11] These techniques make it impossible to directly sample from the space of possible models.”
Learning to Learn (1998, unknown): “In [Atkeson, 1991; Baxter, 1995a; Caruana, 1996; Friedman, 1994; Hastie and Tibshirani, 1994; Mel, 1996; Moore et aI., 1992; Thrun and O'Sullivan, 1996], various algorithms are described for adjusting a parameterized distance metric.”
On-line regression algorithms for learning mechanical models of robots: A survey (2011, methodology): “Locally Weighted Regression (lwr) methods [4] use Gaussian features like rbfns, but the weights are approximated with the ls method (see Fig.”
Teleoperation with intelligent and customizable interfaces (2013, methodology): “A significant amount of work in this area (Atkeson, 1991; Kaiser & Dillmann, 1996; Grollman & Jenkins, 2008) has applied function approximation techniques to learn the policy from examples provided by a human or another robot.1 We adopt the same principle, and use an extension to Online Learning…”
Discovering Structure in Multiple Learning Tasks: The TC Algorithm (1996, methodology): “Following ideas presented elsewhere [3, 4, 10, 11, 15, 16, 28], this is done by minimizing the distance between training examples that belong to the same class, while maximizing the distance between training examples with opposite class labels: En(d) = Xx;y xy distd(x ; y) ! min”
Learning Forward Models for the Operational Space Control of Redundant Robots (2010, methodology): “Locally weighted regressions were ﬁrst used by Atkeson (1991) to realise supervised learning on robots.”
Online customization of teleoperation interfaces (2012, background): “A significant amount of work in this area [13]– [15] has applied function approximation techniques to learn the policy from examples provided by a human or another robot.”
Clustering Learning Tasks and the Selective Cross-Task Transfer of Knowledge (1998, methodology): “Following the ideas in [3, 4, 11, 12, 18, 33], the TC algorithm uses a globally weighted version of distEuclid distd(x ; y) = sXi d(i) x(i) y(i) 2: (4) Here d (with d(i) 0 for all i) is a vector of parameters that determine the relative weight of each input dimension.”
Is Learning The n-th Thing Any Easier Than Learning The First? (1995, background): “An alternative way for exploiting support sets to improve memory-based learning is to learn a distance function [3] [9].”
Pattern recognition and neural networks (1996, methodology): “Like that of Fukunag a & Flick (1984) this is a global method, but Lowe mentions that the feature weights could be chosen locally, as was done by Atkeson (1991) in a control problem.”