Koopa: Learning Non-stationary Time Series Dynamics with Koopman Predictors

Chenyu Li; Jianmin Wang; Mingsheng Long; Yong Liu

Koopa: Learning Non-stationary Time Series Dynamics with Koopman Predictors

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2305.18803 v2 pith:SRHNFTWX submitted 2023-05-30 cs.LG

Koopa: Learning Non-stationary Time Series Dynamics with Koopman Predictors

Yong Liu , Chenyu Li , Jianmin Wang , Mingsheng Long This is my paper

classification cs.LG

keywords koopmandynamicskoopaseriestimenon-stationarytime-variantdeep

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

0 comments

read the original abstract

Real-world time series are characterized by intrinsic non-stationarity that poses a principal challenge for deep forecasting models. While previous models suffer from complicated series variations induced by changing temporal distribution, we tackle non-stationary time series with modern Koopman theory that fundamentally considers the underlying time-variant dynamics. Inspired by Koopman theory of portraying complex dynamical systems, we disentangle time-variant and time-invariant components from intricate non-stationary series by Fourier Filter and design Koopman Predictor to advance respective dynamics forward. Technically, we propose Koopa as a novel Koopman forecaster composed of stackable blocks that learn hierarchical dynamics. Koopa seeks measurement functions for Koopman embedding and utilizes Koopman operators as linear portraits of implicit transition. To cope with time-variant dynamics that exhibits strong locality, Koopa calculates context-aware operators in the temporal neighborhood and is able to utilize incoming ground truth to scale up forecast horizon. Besides, by integrating Koopman Predictors into deep residual structure, we ravel out the binding reconstruction loss in previous Koopman forecasters and achieve end-to-end forecasting objective optimization. Compared with the state-of-the-art model, Koopa achieves competitive performance while saving 77.3% training time and 76.0% memory.

discussion (0)

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Deep Time Series Models: A Comprehensive Survey and Benchmark
cs.LG 2024-07 unverdicted novelty 7.0

This survey and benchmark of deep time series models using the released TSLib library finds that models with specific structures perform well only on distinct analysis tasks.
iTransformer: Inverted Transformers Are Effective for Time Series Forecasting
cs.LG 2023-10 unverdicted novelty 6.0

By applying attention and feed-forward networks to inverted variate tokens instead of temporal tokens, iTransformer achieves state-of-the-art performance on real-world time series forecasting datasets.
Zeus: Towards Tuning-Free Foundation Model for Time Series Analysis
cs.LG 2026-07 unverdicted novelty 5.0

Zeus proposes a multi-scale Transformer with point-wise tokenization and Multi-Objective Temporal Masking to enable tuning-free performance on forecasting, interpolation, and other time series tasks.
Dirichlet-Guided Group Forecasting for Alleviating Over-smoothing in Time Series Forecasting
cs.LG 2026-06 unverdicted novelty 5.0

DGF explicitly models multiple mode-conditioned predictive distributions via Dirichlet-guided sampling and reward optimization to preserve dynamical features in time series forecasts.