1. iTransformer: Inverted Transformers Are Effective for Time Series Forecasting(这篇热度高到爆炸,就不用多说了,分数高是意料之中的)
关键词:时间序列预测,长时预测、Transformer
TL;DR:提出了一种新颖的周期性解耦框架(PDF),通过捕获 2D 时间变化建模来进行长期序列预测。
地址链接:iTransformer: Inverted Transformers Are Effective for Time Series Forecasting | OpenReview
2. ClimODE: Climate Forecasting With Physics-informed Neural ODEs
关键词:神经常微分方程、时间序列预测、气候预测、基于物理的机器学习、不确定性量化
TL;DR:引入了一种受物理学启发的新颖的气候建模方法,使用常微分方程捕获潜在的归纳偏差并允许预测中的不确定性量化。
地址链接:
ClimODE: Climate and Weather Forecasting with Physics-informed Neural ODEs | OpenReview
3. FITS: Modeling Time Series with 10K Parameters
关键词:时间序列分析、时间序列预测、复值神经网络,轻量级
TL;DR:提出了一种新颖的周期性解耦框架(PDF),通过捕获 2D 时间变化建模来进行长期序列预测。
地址链接:
FITS: Modeling Time Series with $10k$ Parameters | OpenReview
4. TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting
关键词:时间序列预测、LLM
TL;DR:这项工作提出了一种用于时间序列预测的GPT
地址链接:
TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting | OpenReview
5. Tim Series Continuous Modeling for Imputation and Forecasting with Implicit Neural Representations
关键词:连续建模、预测、插补、隐式神经表示
TL;DR:提出了一种用于时间序列插补和预测的深度连续模型,可以处理不规则采样时间序列和新样本.
地址链接:
6. TACTiS-2: Better, Faster, Simpler Attentional Copulas for Multivariate Time Series
关键词:多元时间序列预测、copula
TL;DR:一个灵活的多变量概率时间序列预测模型,简化了attentional copulas,在不同的预测任务中具有最先进的精度,同时支持插值和从不规则数据中学习。
地址链接:
TACTiS-2: Better, Faster, Simpler Attentional Copulas for Multivariate Time Series | OpenReview
7. Copula Conformal prediction for multi-step time series prediction
关键词:共形预测、时间序列、不确定性量化、校准、RNN、Copula
TL;DR:通过使用 copula 对时间步长的依赖性进行建模,显着提高共形预测置信区间的效率/清晰度,用于多步时间序列预测。
地址链接:
Copula Conformal prediction for multi-step time series prediction | OpenReview
8. STanHop: Sparse Tandem Hopfield Model for Memory-Enhanced Time Series Prediction
关键词:时间序列预测;多元时间序列;霍普菲尔德网络;
TL,DR:提出了STanHop-Net,一种新颖的时间序列预测模型,将基于 Hopfield 的模块与外部存储模块相结合,增强学习能力,对突发事件快速响应,并具有卓越的理论保证和经验性能。
地址链接:
STanHop: Sparse Tandem Hopfield Model for Memory-Enhanced Time Series Prediction | OpenReview
9. Periodicity Decoupling Framework for Long-term Series Forecasting
关键词:时间序列分解,长时预测
TL;DR:提出了一种新颖的周期性解耦框架(PDF),通过捕获 2D 时间变化建模来进行长期序列预测。
地址链接:
Periodicity Decoupling Framework for Long-term Series Forecasting | OpenReview
10. Self-Supervised Contrastive Forecasting
关键词:自监督,对比学习,长时预测
TL;DR:提出了一种新颖的周期性解耦框架(PDF),通过捕获 2D 时间变化建模来进行长期序列预测。
地址链接:
Self-Supervised Contrastive Learning for Long-term Forecasting | OpenReview
11. TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series
关键词:多元时间序列预测、LLM
地址链接:
TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Series | OpenReview
12. transformer
1.TACTiS-2: Better, Faster, Simpler Attentional Copulas for Multivariate Time Series
2.Periodicity Decoupling Framework for Long-term Series Forecasting
3.Multi-scale Transformers with Adaptive Pathways for Time Series Forecasting
4.VQ-TR: Vector Quantized Attention for Time Series Forecasting
知乎总结:
ICLR 2024 时间序列(Time Series)高分论文
ICLR 2023丨时间序列(Time Series)论文汇总
ICLR 2024丨时间序列(Time Series)论文汇总
精选11篇顶会论文,深度学习时间序列预测模型汇总!(含2024最新)
原理+论文+实战:60篇由浅入深的时间序列预测/分类教程汇总