A Convex Relaxation Approach for Learning the Robust Koopman Operator

Table 1

Average run time for 600 epochs training cases with different bottlenecks. In general, our model is almost two times faster than consistent Koopman AE during training and similar to Koopman AE.

Scenario
Koopman AE	61.18 s	62.38 s	63.49 s
Stiefel Koopman	61.34 s	62.41 s	63.34 s
Consistent Koopman AE	127.44 s	136.27 s	150.85 s