search

EthanJamesLew / AutoKoopman

AutoKoopman - automated Koopman operator methods for data-driven dynamical systems analysis and control.
GNU General Public License v3.0
57 stars 10 forks source link

Obs type "quadratic" throws error if sampling period is too large for "bio2" benchmark #13

Closed Abdu-Hekal closed 2 years ago

Abdu-Hekal commented 2 years ago

For "larger" sampling periods in the "bio2" benchmark (sampling period >= 0.1), the returned "trajectory.states" from "model.solve_vp" returns rows with nan, such that the trajectory returned is incomplete. Whilst running it also throws: "RuntimeWarning: overflow encountered in double_scalars".

Could this be an inherit limitation of the obs type such that for more complex systems it requires a smaller sampling rate?

EthanJamesLew commented 2 years ago

@Abdu-Hekal could you put a code snippet in this issue that reproduces this error? Otherwise, could you email me what you have?

EthanJamesLew commented 2 years ago

Reducing the maximum rank seems to stabilize it. Given that the optimizer chooses the overflowed observables mapping, I will need to change the AutoKoopman hyperparameter optimizer to change this.

EthanJamesLew commented 2 years ago

We will not be using quadratic now that we have a decent polynomial. Closing for now