A unified online algorithm for linear dynamical systems achieves sublinear regret with O~(k) learnable parameters where k measures instability complexity, supported by a lower bound requiring at least k filters for any predictor.
Proceedings of the 24th Annual Conference on Learning Theory , pages =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Memory Efficient Unified Algorithm for Online Learning of Linear Dynamical Systems
A unified online algorithm for linear dynamical systems achieves sublinear regret with O~(k) learnable parameters where k measures instability complexity, supported by a lower bound requiring at least k filters for any predictor.