Universal Approximation and Learning of Trajectories Using Oscillators
Abstract
Natural and artificial neural circuits must be capable of travers(cid:173) ing specific state space trajectories. A natural approach to this problem is to learn the relevant trajectories from examples. Un(cid:173) fortunately, gradient descent learning of complex trajectories in amorphous networks is unsuccessful. We suggest a possible ap(cid:173) proach where trajectories are realized by combining simple oscil(cid:173) lators, in various modular ways. We contrast two regimes of fast and slow oscillations. In all cases, we show that banks of oscillators with bounded frequencies have universal approximation properties. Open questions are also discussed briefly.
Cite
Text
Baldi and Hornik. "Universal Approximation and Learning of Trajectories Using Oscillators." Neural Information Processing Systems, 1995.Markdown
[Baldi and Hornik. "Universal Approximation and Learning of Trajectories Using Oscillators." Neural Information Processing Systems, 1995.](https://mlanthology.org/neurips/1995/baldi1995neurips-universal/)BibTeX
@inproceedings{baldi1995neurips-universal,
title = {{Universal Approximation and Learning of Trajectories Using Oscillators}},
author = {Baldi, Pierre and Hornik, Kurt},
booktitle = {Neural Information Processing Systems},
year = {1995},
pages = {451-457},
url = {https://mlanthology.org/neurips/1995/baldi1995neurips-universal/}
}