A convex spectral learner followed by spectral-to-LDS distillation extracts compact linear state-space predictors from nonlinear dynamics with a provable one-step error decomposition.
SpectraLDS: Provable Distillation for Linear Dynamical Systems
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
We present the first provable method for identifying symmetric linear dynamical systems (LDS) with accuracy guarantees that are independent of the systems' state dimension or effective memory. Our approach builds upon recent work that represents symmetric LDSs as convolutions learnable via fixed spectral transformations. We show how to invert this representation, thereby recovering an LDS model from its spectral transform and yielding an end-to-end convex optimization procedure. This distillation preserves predictive accuracy while enabling constant-time and constant-space inference per token, independent of sequence length. We evaluate our method, SpectraLDS, as a component in sequence prediction architectures and demonstrate that accuracy is preserved while inference efficiency is improved on tasks such as language modeling.
citation-role summary
citation-polarity summary
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Spectral Distillation: From Nonlinear Dynamics to Linear State-Space Models
A convex spectral learner followed by spectral-to-LDS distillation extracts compact linear state-space predictors from nonlinear dynamics with a provable one-step error decomposition.