## Assessing the quality of learned local models (1994)

Venue: | Advances in Neural Information Processing Systems 6 |

Citations: | 43 - 15 self |

### BibTeX

@INPROCEEDINGS{Schaal94assessingthe,

author = {Stefan Schaal and Christopher G. Atkeson},

title = {Assessing the quality of learned local models},

booktitle = {Advances in Neural Information Processing Systems 6},

year = {1994},

publisher = {Morgan Kaufmann}

}

### OpenURL

### Abstract

An approach is presented to learning high dimensional functions in the case where the learning algorithm can affect the generation of new data. A local modeling algorithm, locally weighted regression, is used to represent the learned function. Architectural parameters of the approach, such as distance metrics, are also localized and become a function of the query point instead of being global. Statistical tests are given for when a local model is good enough and sampling should be moved to a new area. Our methods explicitly deal with the case where prediction accuracy requirements exist during exploration: By gradually shifting a “center of exploration ” and controlling the speed of the shift with local prediction accuracy, a goal-directed exploration of state space takes place along the fringes of the current data support until the task goal is achieved. We illustrate this approach with simulation results and results from a real robot learning a complex juggling task. 1

### Citations

1320 | Generalized Additive Models
- Hastie, Tibshirani
- 1990
(Show Context)
Citation Context ...) can not only be used as a diagnostic tool, but they can also serve to optimize the architectural parameters of LWR. This results in a function fitting technique called supersmoothing in statistics (=-=Hastie & Tibshirani, 1991-=-). Fan&Gijbels (1992) investigated a method for this purpose that required estimation of the second derivative of the function to be approximated and its density distribution. These two measures are n... |

114 | Smoothing Techniques with implementation in S - Hardle - 1991 |

101 | Hoeffding races: Accelerating model selection search for classification and function approximation - Maron, Moore - 1994 |

92 | Robot juggling: an implementation of memory-based learning. Control systems magazine, pages 57–71. constraints and intrinsic motivation 65 - Schaal, Atkeson - 1994 |

81 | Variable bandwidth and local linear regression smoothers - Fan, Gijbels - 1992 |

55 | The Computation and Theory of Optimal Control - Dyer, McReynolds - 1970 |

44 |
Memory-based approaches to approximating continuous functions”, in Nonlinear Modeling and
- Atkeson
- 1992
(Show Context)
Citation Context ...duce too much bias while the latter require fitting many parameters which is computationally expensive and needs a lot of data. The algorithm which we explore here, locally weighted regression (LWR) (=-=Atkeson, 1992-=-, Moore, 1991, Schaal&Atkeson, 1994), is closely related to versions suggested by Cleveland et al. (1979, 1988) and Farmer&Siderowich (1987). A LWR model is trained by simply storing every experience ... |

41 | Regression by local fitting: methods, properties and computational algorithms - Cleveland, Devlin, et al. - 1988 |

40 | Regression by Local Fitting - Cleveland, Delvin, et al. - 1988 |

37 |
Robust locally-weighted regression and smoothing scatterplots
- Cleveland
- 1979
(Show Context)
Citation Context ...at every query point instead of representing it in a piecewise linear fashion. The algorithm applied for this purpose, locally weighted regression (LWR), stems from nonparametric regression analysis (=-=Cleveland, 1979-=-, Mller, 1988, Hrdle 1990, Hastie&Tibshirani, 1991). In Section 2, we will briefly outline LWR. Section 3 discusses 2 several statistical tools for assessing the quality of a learned linear LWR model,... |

36 |
Nonparametric regression analysis of longitudinal data.Lecture
- Müller
- 1988
(Show Context)
Citation Context ...int instead of representing it in a piecewise linear fashion. The algorithm applied for this purpose, locally weighted regression (LWR), stems from nonparametric regression analysis (Cleveland, 1979, =-=Müller, 1988-=-, Härdle 1990, Hastie&Tibshirani, 1991). In Section 2, we will briefly outline LWR. Section 3 discussesseveral statistical tools for assessing the quality of a learned linear LWR model, how to optimi... |

15 | Predicting chaotic dynamics - Farmer, Sidorovich - 1988 |

4 | Smoothing Techniques with Implementation in - Hrdle - 1991 |

3 |
Nonparametric regression analysis for longitudinal data
- M��ller
- 1988
(Show Context)
Citation Context ...int instead of representing it in a piecewise linear fashion. The algorithm applied for this purpose, locally weighted regression (LWR), stems from nonparametric regression analysis (Cleveland, 1979, =-=Mller, 1988-=-, Hrdle 1990, Hastie&Tibshirani, 1991). In Section 2, we will briefly outline LWR. Section 3 discusses 2 several statistical tools for assessing the quality of a learned linear LWR model, how to optim... |

3 | Choosing a subset regression." unpublished paper presented at the annual meeting of the American Statistical Association - Mallows - 1966 |

2 | Choosing a subset regression." unpublished paper presented at the annual meeting - Mallows - 1966 |