Koopman Operator Identification of Model Parameter Trajectories for Temporal Domain Generalization (KOMET)

arXiv stat.ML / 3/31/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces KOMET, a model-agnostic framework for temporal domain generalization that views a sequence of trained model parameters as a nonlinear dynamical system and identifies a governing linear operator via EDMD.
It improves robustness to temporal domain drift using a warm-start sequential training protocol that enforces smooth parameter trajectories.
KOMET uses a Fourier-augmented observable dictionary to capture periodic components often present in real-world distribution changes.
After learning the Koopman operator, it predicts future parameter trajectories autonomously at deployment without needing future labeled data, enabling zero-retraining adaptation.
Experiments on six datasets show near-perfect mean autonomous rollout accuracy (0.981–1.000) over 100 held-out time steps, with spectral/coupling analyses indicating interpretable dynamical structure tied to drifting decision-boundary geometry.

Abstract

Parametric models deployed in non-stationary environments degrade as the underlying data distribution evolves over time (a phenomenon known as temporal domain drift). In the current work, we present KOMET (Koopman Operator identification of Model parameter Evolution under Temporal drift), a model-agnostic, data-driven framework that treats the sequence of trained parameter vectors as the trajectory of a nonlinear dynamical system and identifies its governing linear operator via Extended Dynamic Mode Decomposition (EDMD). A warm-start sequential training protocol enforces parameter-trajectory smoothness, and a Fourier-augmented observable dictionary exploits the periodic structure inherent in many real-world distribution drifts. Once identified, KOMET's Koopman operator predicts future parameter trajectories autonomously, without access to future labeled data, enabling zero-retraining adaptation at deployment. Evaluated on six datasets spanning rotating, oscillating, and expanding distribution geometries, KOMET achieves mean autonomous-rollout accuracies between 0.981 and 1.000 over 100 held-out time steps. Spectral and coupling analyses further reveal interpretable dynamical structure consistent with the geometry of the drifting decision boundary.