Alessandra Celletti

Universit√† degli Studi di Roma “Tor Vergata”

From chaos to invariant tori in rotational dynamics: from machine learning to computer-assisted results

We use a simple problem of rotational dynamics, the spin-orbit problem, as a bench-test of different techniques to unveil the regular or chaotic character of the dynamics.
In the first part, we implement different forms of deep learning to classify the motions starting from time series and without any prior knowledge of the underlying dynamics (joint work with C. Gales, V. Rodriguez, M. Vasile).
In the second part, we generalize the model by including also dissipative effects due to the tidal torque and we construct invariant tori by means of a KAM theorem for conformally symplectic systems. The existence of the tori is validated by a computer-assisted technique (joint work with R. Calleja, J.Gimeno, R. de la Llave).
We will highlight the potentials and the need of further developments in machine learning approaches and computer-assisted methods.

Event Timeslots (1)

Tuesday
-
From chaos to invariant tori in rotational dynamics: from machine learning to computer-assisted results
tmscm tmscm