2022人工智能顶会时间序列论文汇总
数据STUDIO转载|作者 贾子钰
新加坡国立大学|研究方向 时序预测
ICML 2022 共收到5630 投稿,接收1117 篇 short oral,118篇 long oral,录用率为21.94%。
整理了ICML 2022关于时间序列相关的论文,围绕时间序列预测、分类、异常检测、表示学习以及在医疗、生物、交通、音乐、金融等方向的应用。01
论文标题: FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting
论文链接:
https://arxiv.org/abs/2201.12740
PPT链接:
https://icml.cc/media/icml-2022/Slides/17986.pdf
代码链接:****https://github.com/MAZiqing/FEDformer
研究方向:
长时间序列预测
02
论文标题:Modeling Irregular Time Series with Continuous Recurrent Units
论文链接:
https://arxiv.org/abs/2111.11344
PPT链接:
https://icml.cc/media/icml-2022/Slides/16343.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/5b4130c9e891d39891289001cc97d86b.png
研究方向:
不规则采样的时间序列建模
03
论文标题:TACTiS: Transformer-Attentional Copulas for Time Series
论文链接:
https://arxiv.org/abs/2202.03528
PPT链接:
https://icml.cc/media/icml-2022/Slides/16594.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/514f94b7b871de0eacb221709d341aec.png
研究方向:
时间序列预测
04
论文标题:Closed-Form Diffeomorphic Transformations for Time Series Alignment
论文链接:
https://arxiv.org/abs/2206.08107
PPT链接:
https://icml.cc/media/icml-2022/Slides/17697.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/371bce7dc83817b7893bcdeed13799b5_Vdn44Un.png
05
论文标题:Reconstructing Nonlinear Dynamical Systems from Multi-Modal Time Series
论文链接:
https://proceedings.mlr.press/v162/kramer22a.html
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/860052df4915de4d6c3deac9f7ebf5cc_I9JU7vo.png
研究方向:
多模态时间序列分析
06
论文标题:Unsupervised Time-Series Representation Learning with Iterative Bilinear Temporal-Spectral Fusion
论文链接:
https://arxiv.org/abs/2202.04770
PPT链接:
https://icml.cc/media/icml-2022/Slides/16051.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/009c434cab57de48a31f6b669e7ba266_hPcNn6H.png
研究方向:
无监督/自监督时间序列(分类、预测和异常检测)
07
论文标题:Deep Variational Graph Convolutional Recurrent Network for Multivariate Time Series Anomaly Detection
论文链接:
https://proceedings.mlr.press/v162/chen22x.html
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/a9be4c2a4041cadbf9d61ae16dd1389e.png
研究方向:
多变量时间序列异常检测
08
论文标题:Learning of Cluster-based Feature Importance for Electronic Health Record Time-series
论文链接:
https://proceedings.mlr.press/v162/aguiar22a.html
PPT链接:
https://icml.cc/media/icml-2022/Slides/17712.pdf
研究方向:
电子健康记录(EHR)数据分析
09
论文标题:Domain Adaptation for Time Series Forecasting via Attention Sharing
论文链接:
https://proceedings.mlr.press/v162/jin22d.html
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/334467d41d5cf21e234465a1530ba647.png
研究方向:
基于DA的时间序列预测
10
论文标题:Adaptive Conformal Predictions for Time Series
论文链接:
https://proceedings.mlr.press/v162/zaffran22a.html
PPT链接:
https://icml.cc/media/icml-2022/Slides/17817_45KQty1.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/b59442085644532ef03417a3e5a76437_ShcNJIg.png
研究方向:
时间序列预测
11
论文标题:Utilizing Expert Features for Contrastive Learning of Time-Series Representations
论文链接:
https://arxiv.org/abs/2206.11517
PPT链接:
https://icml.cc/media/icml-2022/Slides/18038.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/14db62200d8bf46551aa214accafe1df.png
研究方向:
时间序列对比学习
12
论文标题:Transformer Neural Processes: Uncertainty-Aware Meta Learning Via Sequence Modeling
论文链接:
https://arxiv.org/abs/2207.04179
PPT链接:
https://icml.cc/media/icml-2022/Slides/17524_JaSq5IX.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/bf8dd8c68d02e161c28dc9ea139d4784_8KSxYQh.png
研究方向:
序列建模、神经过程、不确定性
一句话总结全文:
提出了变压器神经过程——TNPs,它将不确定性感知元学习视为序列建模问题。
13
论文标题:Accelerating Bayesian Optimization for Biological Sequence Design with Denoising Autoencoders
论文链接:
https://proceedings.mlr.press/v162/stanton22a.html
PPT链接:
https://icml.cc/media/icml-2022/Slides/18028.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/34ad9bc83e3c72c62281cb2c744ac966_McpNQsy.png
研究方向:
生物序列数据建模、贝叶斯优化
14
论文标题:CITRIS: Causal Identifiability from Temporal Intervened Sequences
论文链接:
https://arxiv.org/abs/2202.03169
PPT链接:
https://icml.cc/media/icml-2022/Slides/17426_PzSasFh.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/43cca4b3de2097b9558efefd0ecc3588_USh52N9.png
研究方向:
时间序列因果分析
15
论文标题:Unsupervised Flow-Aligned Sequence-to-Sequence Learning for Video Restoration
论文链接:
https://proceedings.mlr.press/v162/lin22d.html
PPT链接:
https://icml.cc/media/icml-2022/Slides/16143.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/6d70cb65d15211726dcce4c0e971e21c_54IRo89.png
代码链接:
https://github.com/linjing7/VR-Baseline
研究方向:
Sequence-to-Sequence计算机视觉模型
16
论文标题:Proximal Exploration for Model-guided Protein Sequence Design
PPT链接:
https://icml.cc/media/icml-2022/Slides/17764_tb7HE1c.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/9f96f36b7aae3b1ff847c26ac94c604e.png
研究方向:
蛋白质序列分析
17
论文标题:Biological Sequence Design with GFlowNets
PPT链接:
https://icml.cc/media/icml-2022/Slides/17338.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/2327fdecafc97928d5ba62af00a05704_nAgr1l8.png
研究方向:
生物序列分析
18
论文标题:Causal Conceptions of Fairness and their Consequences
论文链接:
https://proceedings.mlr.press/v162/nilforoshan22a.html
PPT链接:
https://icml.cc/media/icml-2022/Slides/17122_eHjIkLG.pdf
研究方向:
因果关系、公平决策算法
19
论文标题:OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework
论文链接:
https://arxiv.org/abs/2202.03052
PPT链接:
https://icml.cc/media/icml-2022/Slides/17835_b3yKPXu.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/6917ff2a7b53421ff4066020e2d89eec_ODAb1me.png
代码链接:
https://github.com/OFA-Sys/OFA
研究方向:
基于Sequence-to-Sequence多模态预训练
20
论文标题:SkexGen: Autoregressive Generation of CAD Construction Sequences with Disentangled Codebooks
论文链接:
https://arxiv.org/abs/2207.04632
PPT链接:
https://icml.cc/media/icml-2022/Slides/18218.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/b6f0479ae87d244975439c6124592772_ksZ7D4K.png
代码链接:
https://samxuxiang.github.io/skexgen
研究方向:
计算机辅助设计(CAD)结构序列分析
21
论文标题:Style Equalization: Unsupervised Learning of Controllable Generative Sequence Models
论文链接:
https://arxiv.org/abs/2110.02891
PPT链接:
https://icml.cc/media/icml-2022/Slides/16200.pdf
海报链接:
https://icml.cc/media/PosterPDFs/ICML%202022/7e6ff0205749bc6025b51155e26f6ced.png
研究方向:
可控序列生成、无监督
来源:https://zhuanlan.zhihu.com/p/553385328 作者:贾子钰
🏴☠️宝藏级🏴☠️ 原创公众号『 数据STUDIO 』内容超级硬核。公众号以Python为核心语言,垂直于数据科学领域,包括 可戳 👉 Python | MySQL | 数据分析 | 数据可视化 | 机器学习与数据挖掘 | 爬虫 等,从入门到进阶!
长按👇关注- 数据STUDIO -设为星标,干货速递