Associate Radial Basis Function Networks

Betreuerin: Rania Rayyes

The associative dynamic network allows to learn redundant multiple
solutions (e.g., learning kinematics of a high DoF robot) by means of
multi-stable attractor dynamics.

Offline and online incremental associate RBF have been already proposed
as associative dynamic networks

In this seminar, the student is expected to understand:

- what the associative dynamic network is and how it works

- comparison between an offline and an online versions.

- an example of its application in robotics.

R. Rayyes and J. Steil, "Online Associative Multi-Stage Goal Babbling Toward Versatile Learning of Sensorimotor Skills," 2019 Joint IEEE 9th International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob), Oslo, Norway, 2019, pp. 327-334.
https://ieeexplore.ieee.org/abstract/document/8850707

R. F. Reinhart, J. J. Steil, "Learning whole upper body control with dynamic redundancy resolution in coupled associative radial basis function networks", IEEE/RSJ IROS, 2012.

https://ieeexplore.ieee.org/document/6385873

R. Reinhart, M. Rolf, "Learning versatile sensorimotor coordination with goal babbling and neural associative dynamics", IEEE ICDL, pp. 1-7, 2013.

https://ieeexplore.ieee.org/document/6652566