ICLR 2025 | 时间序列(Time Series)高分论文汇总

ICLR2025已经结束了讨论阶段，进入了meta-review阶段，分数应该不会有太大的变化了，本文总结了其中时间序列(Time Series)高分的论文。如有疏漏，欢迎大家补充。

挑选原则：均分要大于等于6（即使有3，但是有8或者更高的分拉回来也算）

1 TimeMixer++: A General Time Series Pattern Machine for Universal Predictive Analysis

链接：https://openreview.net/forum?id=1CLzLXSFNn

分数：6810

关键词：多任务（预测，分类，插补，异常检测），基础模型

keywords： time series, pattern machine, predictive analysis

TL; DR ：TimeMixer++ is a time series pattern machine that employs multi-scale and multi-resolution pattern extraction to deliver SOTA performance across 8 diverse analytical tasks, including forecasting, classification, anomaly detection, and imputation.

2 Root Cause Analysis of Anomalies in Multivariate Time Series through Granger Causal Discovery

链接：https://openreview.net/forum?id=k38Th3x4d9

分数：88888

关键词：因果发现

keywords：root cause analysis, Granger causality, multivariate time series

AERCA

3 Amortized Control of Continuous State Space Feynman-Kac Model for Irregular Time Series

链接：https://openreview.net/forum?id=8zJRon6k5v

分数：8888

关键词：变分推断，不规则时间序列，状态空间模型

keywords：stochastic optimal control, variational inference, state space model, irregular time series

TL; DR：We propose a multi-marginal Doob's $h$-transform for irregular time series and variational inference with stochastic optimal control to approximate it.

4 Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts

链接：https://openreview.net/forum?id=e1wDDFmlVu

分数：688

关键词：预测，基础模型，混合专家系统

keywords：time series, foundation model, forecasting

Time-MoE

5 Label Correlation Biases Direct Time Series Forecast

链接：https://openreview.net/forum?id=4A9IdSa1ul

分数：8686

关键词：长时预测，频域

keywords：Time series, Long-term Forecast

TL; DR：Learning to forecast in the frequency domain significantly enhances forecasting performance.

6 Fast and Slow Streams for Online Time Series Forecasting Without Information Leakage

链接：https://openreview.net/forum?id=I0n3EyogMi

分数：6688

关键词：在线预测，流式数据，概念飘逸

keywords：online time series forecasting, concept drift, online learning

TL; DR： Redefined the setting of online time series forecasting to prevent information leakage and proposed a model-agnostic framework for this setting.

7 Shifting the Paradigm: A Diffeomorphism Between Time Series Data Manifolds for Achieving Shift-Invariancy in Deep Learning

链接：https://openreview.net/forum?id=nibeaHUEJx

分数：6688

关键词：频域，平移不变性

keywords：Time series analysis, invariance in neural networks

8 Optimal Transport for Time Series Imputation

链接：https://openreview.net/forum?id=xPTzjpIQNp

分数：588

关键词：插补，最优传输

keywords： Time series, Imputation

9 Constrained Posterior Sampling: Time Series Generation with Hard Constraints

链接：https://openreview.net/forum?id=pKMpmbuKnd

分数：5688

关键词：时间序列生成，扩散模型

keywords：Time Series Generation, Posterior Sampling, Diffusion Models, Controlled Generation

10 A Simple Baseline for Multivariate Time Series Forecasting

链接：https://openreview.net/forum?id=oANkBaVci5

分数：5688

关键词：预测，小波变换

keywords：Time Series Forecasting, Wavelets

11 Shedding Light on Time Series Classification using Interpretability Gated Networks

链接：https://openreview.net/forum?id=n34taxF0TC

分数：56688

关键词：可解释性，Shapelet（特征提取）

keywords：Interpretability, Time-series, Shapelet

TL; DR： A framework to integrate interpretable models with deep neural networks for interpretable time-series classification.

12 Multi-Resolution Decomposable Diffusion Model for Non-Stationary Time Series Anomaly Detection

链接：https://openreview.net/forum?id=eWocmTQn7H

分数：6668

关键词：异常检测，多分辨率，扩散模型

keywords：Diffusion Model, Non-Stationary Time Series, Anomaly Detection, Multi-Resolution

TL; DR：This paper delves into the potential of multi-resolution technique and diffusion model for non-stationary time series anomaly detection, supported by rigorous mathematical proofs.

13 CATCH: Channel-Aware Multivariate Time Series Anomaly Detection via Frequency Patching

链接：https://openreview.net/forum?id=m08aK3xxdJ

分数：5668

关键词：异常检测，频域

keywords： Multivariate Time Series, Anomaly Detection

CATCH

14 CoMRes: Semi-Supervised Time Series Forecasting Utilizing Consensus Promotion of Multi-Resolution

链接：https://openreview.net/forum?id=bRa4JLPzii

分数：5668

关键词：多尺度，半监督

keywords：Time series forecasting, Multi-scale, Semi-supervised learning

TL; DR：we propose a novel semi-supervised time series forecasting utilzing con

15 Towards Neural Scaling Laws for Time Series Foundation Models

链接：https://openreview.net/forum?id=uCqxDfLYrB

分数：5668

keywords：Time series, scaling law, foundation model, transformer, forecasting

16 Quantifying Past Error Matters: Conformal Inference for Time Series

链接：https://openreview.net/forum?id=RD9q5vEe1Q

分数：5668

关键词：不确定性量化，分布偏移

keywords：Time Series; Uncertainty Quantification; Conformal Prediction; Distribution Shift

17 TVNet: A Novel Time Series Analysis Method Based on Dynamic Convolution and 3D-Variation

链接：https://openreview.net/forum?id=MZDdTzN6Cy

分数：5668

关键词：卷积

keywords：Time series Analysis, Dynamic convolution, Deep Learning

TL; DR：New time series modeling perspective based 3D-variation and new analysis framework based dynamic convolution

18 In-context Time Series Predictor

链接：https://openreview.net/forum?id=dCcY2pyNIO

分数：3668

关键词：预测，上下文学习

keywords：Time Series Forecasting, In-context Learning, Transformer

19 Compositional simulation-based inference for time series

链接：https://openreview.net/forum?id=uClUUJk05H

分数：566668

关键词：贝叶斯推断

keywords：Simulation-based inference, Bayesian inference, time series, markovian simulators, Amortized Bayesian inference

FNSE

20 Towards a General Time Series Anomaly Detector with Adaptive Bottlenecks and Dual Adversarial Decoders

链接：https://openreview.net/forum?id=aKcd7ImG5e

分数：6666

关键词：异常检测

keywords：Time series, Anomaly detection

TL; DR：We propose a general time series anomaly detection model that is pre-trained on multi-domain datasets and can subsequently apply to many downstream scenarios

21 TimeKAN: KAN-based Frequency Decomposition Learning Architecture for Long-term Time Series Forecasting

链接：https://openreview.net/forum?id=wTLc79YNbh

分数：3588

关键词：预测，KAN

keywords：kolmogorov-Arnold Network; Time Series Forecasting

TimeKAN

22 Investigating Pattern Neurons in Urban Time Series Forecasting

链接：https://openreview.net/forum?id=a9vey6B54y

分数：6666

关键词：时空预测（更像是），城市时间序列预测模型

keywords：urban time series forecasting, neuron detection

PN-Train

23 Locally Connected Echo State Networks for Time Series Forecasting

链接：https://openreview.net/forum?id=KeRwLLwZaw

分数：6666

关键词：回声状态网络

keywords：Time Series Analysis, Time Series Forecasting, Recurrent Networks, Regression, Echo State Networks

TL; DR： Improved locally connected ESN method comparable with state-of-the-art on real-world time series datasets.

24 Diffusion-based Decoupled Deterministic and Uncertain Framework for Probabilistic Multivariate Time Series Forecasting

链接：https://openreview.net/forum?id=HdUkF1Qk7g

分数：6666

关键词：长时预测，扩散模型

keywords：long-term time series forecasting, deep learning, diffusion model

D^3U

25 TS-LIF: A Temporal Segment Spiking Neuron Network for Time Series Forecasting

链接：https://openreview.net/forum?id=rDe9yQQYKt

分数：666

关键词：脉冲神经网络

keywords：spiking neural network, time series forecasting, Application

TL; DR：We proposed a Temporal Segment Spiking Neuron Network (TS-LIF) for multivariate time series forecasting, supported by stability analysis and frequency response analysis to demonstrate its effectiveness and efficiency.

26 Exploring Representations and Interventions in Time Series Foundation Models

链接：https://openreview.net/forum?id=IRL9wUiwab

分数：6666

keywords：Time Series Foundation Models, Model Steering, Interpretability, Pruning

TL; DR：We investigate why time series foundation models work, the kinds of concepts that these models learn, and how can these concepts be manipulated to influence their outputs?

27 FLDmamba: Integrating Fourier and Laplace Transform Decomposition with Mamba for Enhanced Time Series Prediction

链接：https://openreview.net/forum?id=9EiWIyJMNi

分数：556668

关键词：Mamba,FFT

keywords：Mamba; Time Series Prediction

FLDmamba

28 KooNPro: A Variance-Aware Koopman Probabilistic Model Enhanced by Neural Processes for Time Series Forecasting

链接：https://openreview.net/forum?id=5oSUgTzs8Y

分数：66666

keywords：Probabilistic time series prediction; Neural Process; Deep Koopman model

29 Context-Alignment: Activating and Enhancing LLMs Capabilities in Time Series

链接：https://openreview.net/forum?id=syC2764fPc

分数：6666

keywords：Time Series, Large Language Models, Context-Alignment

TL; DR：LLMs for time series tasks

30 TwinsFormer: Revisiting Inherent Dependencies via Two Interactive Components for Time Series Forecasting

链接：https://openreview.net/forum?id=BSsyY29bcl

分数：55568

keywords：Inherent Dependencies, Interactive Components, Time Series Forecasting

TL; DR：A novel Transformer-and decomposition-based framework using residual and interactive learning for time series forecasting.

31 DyCAST: Learning Dynamic Causal Structure from Time Series

链接：https://openreview.net/forum?id=WjDjem8mWE

分数：3668

关键词：

TL; DR：dynamic causal discovery; time series

32 Drift2Matrix: Kernel-Induced Self Representation for Concept Drift Adaptation in Co-evolving Time Series

链接：https://openreview.net/forum?id=prSJlvWrgE

分数：3866

TL; DR：co-evolving time series, concept drift, kernel representation learning

 时间序列领域做毕设！论文方向没灵感没创新点？Informer+Timesnet代码复现，80篇＋前沿论文，110篇顶会顶刊原文/演示应有尽有！来领一份。