Reference Type  Conference Proceedings 
Author(s)  Ijspeert, A.;Nakanishi, J.;Schaal, S. 
Year  2003 
Title  Learning attractor landscapes for learning motor primitives 
Journal/Conference/Book Title  Advances in Neural Information Processing Systems 15 
Keywords  learning nonlinear attractor landscapes
movement primitives
humanoid robotics
statistical learning 
Abstract  If globally high dimensional data has locally only low dimensional distributions, it is advantageous to perform a local dimensionality reduction before further processing the data. In this paper we examine several techniques for local dimensionality reduction in the context of locally weighted linear regression. As possible candidates, we derive local versions of factor analysis regression, principle component regression, principle component regression on joint distributions, and partial least squares regression. After outlining the statistical bases of these methods, we perform Monte Carlo simulations to evaluate their robustness with respect to violations of their statistical assumptions. One surprising outcome is that locally weighted partial least squares regression offers the best average results, thus outperforming even factor analysis, the theoretically most appealing of our candidate techniques.Ê 
Editor(s)  Becker, S.;Thrun, S.;Obermayer, K. 
Publisher  Cambridge, MA: MIT Press 
Pages  15471554 
Short Title  Learning attractor landscapes for learning motor primitives 
URL(s)  http://wwwclmc.usc.edu/publications/I/ijspeertNIPS2002.pdf

Research Notes  ¥ The authors model discrete and rhythmic movement with nonlinear differential equation that form stable point and limit cycle attractors, respectively. The novel finding of this modeling approach was that such equation can be trivially learned with optim 