AAAI 2026 | 时间序列(Time Series) 论文总结[下] (分类，异常检测，基础模型，表示学习，生成等)

时空探索之旅

发布于 2026-03-10 14:42:37

1090

文章被收录于专栏：时空探索之旅时空探索之旅

AAAI 2026将在2026年1月20日到1月27日于新加坡（Singapore）举行。AAAI 2026会议主会共有23, 680篇论文投稿，其中4, 167 篇被接收，接收率为17.6%。

本文总结了2026 AAAI 上有关时间序列（Time Series）相关论文。如有疏漏，欢迎大家补充。

时间序列Topic：时间序列预测，异常检测，分类，生成，表示学习，医疗时序，基础模型，大语言模型等内容。总计67篇，本文涉及32篇。

注：由于论文数目较多，分为上下篇，此为下篇，主要涵盖分类，异常检测，基础模型，表示学习，生成等。

36. A Unified Shape-Aware Foundation Model for Time Series Classification37. MedSpaformer: A Transferable Transformer with Multi-Granularity Token Sparsification for Medical Time Series Classification38. Counterfactual eXplainable AI (XAI) Method for Deep Learning-Based Multivariate Time Series Classification39. EdgeMTSC: A Lightweight Large-Kernel ConvNet for Multivariate Time Series Classification40. FlowPath: Learning Data-Driven Manifolds with Invertible Flows for Robust Irregularly-sampled Time Series Classification41. SVGL: Scale-Variable Graph Learning in Model Space for Multivariate Time Series Classification42. CANDI: Curated Test-Time Adaptation for Multivariate Time-Series Anomaly Detection Under Distribution Shift43. Finding Time Series Anomalies Using Granular-Ball Vector Data Description44. DoKnowAD: Calibrating Normal Representations with Refined Domain Knowledge to Enhance Time Series Anomaly Detection45. Harnessing Vision-Language Models for Time Series Anomaly Detection46. A Theoretical Analysis of Detecting Large Model-Generated Time Series47. State-Derivative-Aware Neural Controlled Differential Equations for Multivariate Time Series Anomaly Detection and Diagnosis48. ORTCL: Towards Continual Learning of Time Series Foundation Models on Streaming Data via Orthogonal Rotation49. DeepSenseMoE: Harnessing Power of Time Series Foundation Models for Few-Shot Human Activity Recognition50. FreqCycle: A Multi-Scale Time-Frequency Analysis Method for Time Series Forecasting51. Koopman Invariants as Drivers of Emergent Time-Series Clustering in Joint-Embedding Predictive Architectures52. Mask the Redundancy: Evolving Masking Representation Learning for Multivariate Time-Series Clustering53. COGS: A Causal Representation Learning Framework for Out-of-Distribution Generalization in Time Series54. Role Hypergraph Contrastive Learning for Multivariate Time-Series Analysis55. Intervention-Aware Time Series Modeling: Capturing and Evaluating Feature Dependencies56. Unifying Channel Independence and Mixing: Multi-Scale Patch Recursion for Global-local Representation Synergy in Multivariate Time Series Forecasting57. Beyond Observations: Reconstruction Error-Guided Irregularly Sampled Time Series Representation Learning58. TGCD: A Framework for Generalized Category Discovery in Time-Series Data59. CaT-Diff: Cascaded Text-enhanced Diffusion Model for Time-Series Imputation60. Dynamic Semantic Tokenization for Time Series via Elastic Sampling on Physics-aware Perception61. TSGDiff: Rethinking Synthetic Time Series Generation from a Pure Graph Perspective62. Neural Architecture and Hyperparameter Selection Through Meta-Learning on Time Series63. Time-Frequency Augmented Multi-level Contrastive Clustering for Time Series64. DiM-TS: Bridge the Gap Between Selective State Space Models and Time Series for Generative Modeling65. Time Series Class-Incremental Learning via Confidence-guided Mask Distillation and Prototype-guided Contrastive Learning66. R-Tuning: Wavelet-Decomposed Replay and Semantic Alignment for Continual Adaptation of Pretrained Time-Series Models67. Learnable Matrix Profile for Motif Discovery on Multivariate Time Series

36 A Unified Shape-Aware Foundation Model for Time Series Classification

链接：https://github.com/ZLiu21/UniShape

作者：Zhen Liu; Yucheng Wang; Boyuan Li; Junhao Zheng; Emadeldeen Eldele; Min Wu; Qianli Ma

关键词：分类，基础模型

37 MedSpaformer: A Transferable Transformer with Multi-Granularity Token Sparsification for Medical Time Series Classification

链接：https://arxiv.org/abs/2503.15578

作者：Jiexia YE; Weiqi Zhang; Ziyue Li; Jia Li; Fugee Tsung

关键词：分类，医疗时序，迁移，多粒度

38 Counterfactual eXplainable AI (XAI) Method for Deep Learning-Based Multivariate Time Series Classification

链接：https://arxiv.org/abs/2511.13237

作者：Alan Gabriel Paredes Cetina; Kaouther Benguessoum; Raoni Lourenco; Sylvain Kubler

关键词：分类，反事实，可解释性AI

39 EdgeMTSC: A Lightweight Large-Kernel ConvNet for Multivariate Time Series Classification

作者：Xueyi Zhou; Zhenyu Li; Dong-Kyu Chae

关键词：分类，轻量化大核卷积

40 FlowPath: Learning Data-Driven Manifolds with Invertible Flows for Robust Irregularly-sampled Time Series Classification

链接：https://www.arxiv.org/abs/2511.10841

作者：YongKyung Oh; Dongyoung Lim; Sungil Kim

关键词：分类，不规则采样

41 SVGL: Scale-Variable Graph Learning in Model Space for Multivariate Time Series Classification

作者：Shikang Liu; Ziyu Tang; Xiren Zhou; Huanhuan Chen

关键词：分类，图学习

42 CANDI: Curated Test-Time Adaptation for Multivariate Time-Series Anomaly Detection Under Distribution Shift

作者：HyunGi Kim; Jisoo Mok; Hyungyu Lee; Juhyeon Shin; Sungroh Yoon

关键词：异常检测，分布偏移，测试时适应

43 Finding Time Series Anomalies Using Granular-Ball Vector Data Description

链接：https://www.arxiv.org/abs/2511.12147

作者：Lifeng Shen; Liang Peng; RuiwenLiu; Shuyin Xia; Y I Liu

关键词：异常检测，聚类

44 DoKnowAD: Calibrating Normal Representations with Refined Domain Knowledge to Enhance Time Series Anomaly Detection

作者：Shiwang Xing; Jianwei Niu; Tao Ren

关键词：异常检测，知识校准

45 Harnessing Vision-Language Models for Time Series Anomaly Detection

链接：https://arxiv.org/abs/2506.06836

作者：Zelin He; Sarah Alnegheimish; Matthew Reimherr

关键词：异常检测，视觉模型

46 A Theoretical Analysis of Detecting Large Model-Generated Time Series

链接：https://arxiv.org/abs/2511.07104

作者：Junji Hou; Junzhou Zhao; Shuo Zhang; Pinghui Wang

关键词：异常检测，时间序列大模型

47 State-Derivative-Aware Neural Controlled Differential Equations for Multivariate Time Series Anomaly Detection and Diagnosis

作者：Xin Sun; Heng Zhou; Yuhao Wu; Chao Li

关键词：异常检测，诊断，微分方程

48 ORTCL: Towards Continual Learning of Time Series Foundation Models on Streaming Data via Orthogonal Rotation

作者：Li Lin; Xinrui Zhang; Qi Zhang; Shuai Wang; Kaiwen Xia

关键词：基础模型，持续学习，流式数据

49 DeepSenseMoE: Harnessing Power of Time Series Foundation Models for Few-Shot Human Activity Recognition

作者：Zenan Fu; Dongzhou Cheng; Lei Zhang; Wenbo Huang; Zhenghao Chen; Hao Wu

关键词：基础模型，少样本人类活动识别

50 FreqCycle: A Multi-Scale Time-Frequency Analysis Method for Time Series Forecasting

作者：Boya Zhang; Shuaijie Yin; Huiwen Zhu; Xing He

关键词：预测，时频分析

51 Koopman Invariants as Drivers of Emergent Time-Series Clustering in Joint-Embedding Predictive Architectures

链接：https://arxiv.org/abs/2511.09783

作者：Pablo Ruiz-Morales; Dries Vanoost; Davy Pissoort; Mathias Verbeke

关键词：聚类，库普曼理论

52 Mask the Redundancy: Evolving Masking Representation Learning for Multivariate Time-Series Clustering

链接：https://arxiv.org/abs/2511.17008

作者：Zexi Tan; Xiaopeng Luo; Yunlin Liu; Yiqun Zhang

关键词：表示学习，聚类

53 COGS: A Causal Representation Learning Framework for Out-of-Distribution Generalization in Time Series

作者：Xinxin Song; Yuxiao Cheng; Tingxiong Xiao; Jinli Suo

关键词：因果表示学习，分布外泛化

54 Role Hypergraph Contrastive Learning for Multivariate Time-Series Analysis

作者：Rundong Xue; Hao Hu; Zhitao Zeng; Xiangmin Han; Zhiqiang Tian; Shaoyi Du; Yue Gao

关键词：时序分析，超图对比学习

55 Intervention-Aware Time Series Modeling: Capturing and Evaluating Feature Dependencies

作者：Ibrahim Delibasoglu; Sanjay Chakraborty; Fredrik Heintz; Mattias Tiger

关键词：时序建模，特征依赖性

56 Unifying Channel Independence and Mixing: Multi-Scale Patch Recursion for Global-local Representation Synergy in Multivariate Time Series Forecasting

作者：Wenhao Zhang; Chun Zhang; Wei Bai; Ning Zhang; Changxia Gao; Yuxin Jia; Chenhao Shi; Shaoxiong Pang

关键词：预测，通道独立，全局-局部

57 Beyond Observations: Reconstruction Error-Guided Irregularly Sampled Time Series Representation Learning

链接：https://arxiv.org/abs/2511.06854

作者：Jiexi Liu; Meng Cao; Songcan Chen

关键词：表示学习，不规则时序

58 TGCD: A Framework for Generalized Category Discovery in Time-Series Data

作者：Chandan Gautam; Lew Choon Hean; Ankit Das; Xiaoli Li; Savitha Ramasamy

关键词：类别发现，泛化

59 CaT-Diff: Cascaded Text-enhanced Diffusion Model for Time-Series Imputation

作者：Changjian Xu; Yong Wang; Ruizheng Huang; Zhicheng Zhang; Wen Yin; Kexin Li

关键词：插补，扩散模型，文本增强

60 Dynamic Semantic Tokenization for Time Series via Elastic Sampling on Physics-aware Perception

作者：Huaizhang Liao; Zhixiong Yang; Jingyuan Xia; Yuheng Sun; Yue Zhang; Shengxi Li; Yongxiang Liu

关键词：时序tokenize，物理感知

61 TSGDiff: Rethinking Synthetic Time Series Generation from a Pure Graph Perspective

链接：https://arxiv.org/abs/2511.12174

作者：Lifeng Shen; Xuyang Li; Lele Long

关键词：生成，图

62 Neural Architecture and Hyperparameter Selection Through Meta-Learning on Time Series

作者：Erfan Moeini; Christopher Vox; Marie Anastacio; Wadie Skaf; Mitra Baratchi; Holger H. Hoos

关键词：神经架构，超参选择，元学习

63 Time-Frequency Augmented Multi-level Contrastive Clustering for Time Series

作者：Congyu Wang; Mingjing Du; Jiang Xiang

关键词：时频增强，对比学习

64 DiM-TS: Bridge the Gap Between Selective State Space Models and Time Series for Generative Modeling

链接：https://arxiv.org/abs/2511.18312

作者：Zihao Yao; Jiankai Zuo; Yaying Zhang

关键词：生成，状态空间模型（Mamba）

65 Time Series Class-Incremental Learning via Confidence-guided Mask Distillation and Prototype-guided Contrastive Learning

作者：Yu Liu; Haoqin Yang; Jinping Sui; Hui Wang; Haipeng Li; Weimin Wang; Qi Jia

关键词：增量学习，对比学习，蒸馏

66 R-Tuning: Wavelet-Decomposed Replay and Semantic Alignment for Continual Adaptation of Pretrained Time-Series Models

链接：https://www.arxiv.org/abs/2511.11685

作者：Tianyi Yin; Jingwei Wang; Chenze Wang; Han Wang; Jiexuan Cai; Min Liu; Yunlong Ma; Kun Gao; Yuting Song; Weiming Shen

关键词：小波分解，语义对齐

67 Learnable Matrix Profile for Motif Discovery on Multivariate Time Series

作者：Mingkai Lin; Yinke Wang; Xiaobin Hong; Wenzhong Li

关键词：模式发现，矩阵概要