Autonomous rollout classification on Dataset F expanding
1Mean Training AccuracyKOMET
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| KOMETDrift type=expanding, Classes / nθ=3/27, Koopman basis=Detrend+Fourier, PCs / nθ=5/27, Spectral radius ρ(A)=1.000, Rollout steps=100, Zero retraining=true2026.03 | 1 | 1 | 1 | 1 | 0 | — | — | — | — | — | — | |
| FrozenDrift type=expanding, Weights=t=299 weights unchanged throughout2026.03 | — | — | — | — | — | — | 0.3314 | 71 | — | — | — | |
| Retrained (upper bound)Drift type=expanding, Mode=Fresh retraining at every prediction step2026.03 | — | — | — | — | — | 1 | — | — | — | — | — |