type
status
date
slug
summary
tags
category
icon
password
阅读概括
1. reader:读者姓名
2. journal/conference:填写期刊或者会议简称及发表时间,如NeurIPS2023
3. title:文章标题
4. author:文章的一作和通讯
5. paper_url:文章的pdf链接
6. code_url:文章若有开源代码,给出开源代码链接
7. task:文章针对的是什么任务,在这些里面选:
1. Classification(分类)
2. Imputation(补全)
3. Prediction (预测)
4. Anomaly detection(异常检测)
5. Clustering (聚类)
6. Compression(压缩)
7. 其它的注明一下
8. domain:文章涉及的是什么领域
1. General(通用领域)
2. Traffic(交通)
3. Healthcare (健康)
4. Finance (金融)
5. Education (教育)
9. challenge:文章要解决的challenge是什么,总结成几条自然语言句子
10. method:文章的方法是基于什么的,在这些里面选:
1. Transformer-based
2. GNN-based
3. Diffusion model-based
4. GAN-based
5. RNN-based
6. TCN-based
7. AE/VAE-based
8. MLP-based
9. 其它的注明一下
11. intuition:用一两句话总结一下作者做这个工作的出发点,想法是什么
12. setting:文章实验的任务设置是什么
1. Supervised learning
2. Unsupervised learning
3. Semi-supervised learning
4. Self-supervised learning
13. improvements:文章的工作相较于以往的工作提升在哪里
14. dataset:文章在哪些dataset上做了实验
TS相关论文收集仓库:
TSFpaper
ddz16 • Updated Jan 13, 2024
2023/11/09
TimesNet (ICLR-2023)2023/11/28
Koopa(NIPS-2023)
2023/12/06
2023/12/13
SAN(NeurIPS 2023)RevIN(ICLR 2022)
论文总结
Titile
Journal/Conference
task
method
domain
challenge
ICLR2023
classification
imputation
prediction/forecasting
anomaly detection
CNN/TCN
general
1. Complex Temporal Patterns: The intermixing and overlapping of various changes (such as rising, falling, fluctuating, etc.) in time series add complexity to modeling efforts.
2. Long-term Dependencies: Existing methods like RNNs, TCNs, and Transformers struggle to effectively capture long-term dependencies in time series.
3. Multi-periodicity: Time series often exhibit multi-periodicity, where the interaction of multiple periods further complicates the modeling process.
- 作者:王大卫
- 链接:https://tangly1024.com/article/essay-time-series
- 声明:本文采用 CC BY-NC-SA 4.0 许可协议,转载请注明出处。