Machine Learning in Celestial Mechanics and Astrodynamics - Prof. Rodriguez-Fernandez

Classification of orbital motions with deep learning and time series data

Prof. Victor Rodriguez-Fernandez

In this presentation we show the capabilities of deep neural networks at classifying types of motion (rotational, librational or chaotic) starting from time series, without any prior knowledge of the underlying dynamics. We will test the ability of different architectures at predicting the character of the dynamics by simply observing a time-ordered set of data. In addition, we will study the generalization power of this approach to predict the character of motions governed by another dynamic model different from the one used for training the machine, and analyse some right and wrong predictions of the network to get insights on its behaviour.

"Within the framework of 'Machine Learning and Computer Assisted
Proofs in Celestial Mechanics and Astrodynamics', June 18 & 25 2021.
https://mlcap2021.wordpress.com/"