Advanced search

Stable Output Feedback in Reservoir Computing Using Ridge Regression

Francis wyffels (UGent) , Benjamin Schrauwen (UGent) and Dirk Stroobandt (UGent)
Author
Organization
Abstract
An important property of Reservoir Computing, and signal processing techniques in general, is generalization and noise robustness. In tra jectory generation tasks, we don't want that a small deviation leads to an instability. For forecasting and system identification we want to avoid over-fitting. In prior work on Reservoir Computing, the addition of noise to the dynamic reservoir tra jectory is generally used. In this work, we show that high-performing reservoirs can be trained using only the commonly used ridge regression. We experimentally validate these claims on two very different tasks: long-term, robust tra jectory generation and system identification of a heating tank with variable dead-time.
Keywords
regularization, reservoir computing

Citation

Please use this url to cite or link to this publication:

Chicago
wyffels, Francis, Benjamin Schrauwen, and Dirk Stroobandt. 2008. “Stable Output Feedback in Reservoir Computing Using Ridge Regression.” In Lecture Notes in Computer Science, ed. V. Kurkova, R. Neruda, and J. Koutnik, 5163:808–817. Berlin: Springer.
APA
wyffels, F., Schrauwen, B., & Stroobandt, D. (2008). Stable Output Feedback in Reservoir Computing Using Ridge Regression. In V. Kurkova, R. Neruda, & J. Koutnik (Eds.), LECTURE NOTES IN COMPUTER SCIENCE (Vol. 5163, pp. 808–817). Presented at the 18th International Conference on Arificial Neural Networks (ICANN 2008), Berlin: Springer.
Vancouver
1.
wyffels F, Schrauwen B, Stroobandt D. Stable Output Feedback in Reservoir Computing Using Ridge Regression. In: Kurkova V, Neruda R, Koutnik J, editors. LECTURE NOTES IN COMPUTER SCIENCE. Berlin: Springer; 2008. p. 808–17.
MLA
wyffels, Francis, Benjamin Schrauwen, and Dirk Stroobandt. “Stable Output Feedback in Reservoir Computing Using Ridge Regression.” Lecture Notes in Computer Science. Ed. V. Kurkova, R. Neruda, & J. Koutnik. Vol. 5163. Berlin: Springer, 2008. 808–817. Print.
@inproceedings{678847,
  abstract     = {An important property of Reservoir Computing, and signal processing techniques in general, is generalization and noise robustness. In tra jectory generation tasks, we don't want that a small deviation leads to an instability. For forecasting and system identi\unmatched{fb01}cation we want to avoid over-\unmatched{fb01}tting. In prior work on Reservoir Computing, the addition of noise to the dynamic reservoir tra jectory is generally used. In this work, we show that high-performing reservoirs can be trained using only the commonly used ridge regression. We experimentally validate these claims on two very di\unmatched{fb00}erent tasks: long-term, robust tra jectory generation and system identi\unmatched{fb01}cation of a heating tank with variable dead-time.},
  author       = {wyffels, Francis and Schrauwen, Benjamin and Stroobandt, Dirk},
  booktitle    = {LECTURE NOTES IN COMPUTER SCIENCE},
  editor       = {Kurkova, V. and Neruda, R. and Koutnik, J.},
  isbn         = {978-3-540-87535-2},
  issn         = {0302-9743},
  keyword      = {regularization,reservoir computing},
  language     = {eng},
  location     = {Prague},
  pages        = {808--817},
  publisher    = {Springer},
  title        = {Stable Output Feedback in Reservoir Computing Using Ridge Regression},
  volume       = {5163},
  year         = {2008},
}

Web of Science
Times cited: