Christian Klos, Yaroslav Felipe Kalle Kossio, Sven Goedeke, Aditya Gilra, and Raoul-Martin Memmesheimer
Neural Network Dynamics and Computation, Institute of Genetics, University of Bonn, Bonn, Germany.
This repository contains example code for the dynamical learning of dynamics using neural networks. The Jupyter notebook creates a network that dynamically learns sinusoidal oscillations with previously unseen period, without changing its weights. See our article in Physical Review Letters (also on the arXiv) for further details and applications. To immediatly run the notebook in your web browser using Colab, click here.
The code was written in Python 3.7.2 using NumPy 1.15.4 and Matplotlib 3.03.