Salient Subsequence Learning for Time Series Clustering论文笔记
2019-06-03 21:36
2106 查看
Salient Subsequence Learning for Time Series Clustering
Qin Zhang, Jia Wu, Member, IEEE, Peng Zhang, Guodong Long, Chengqi Zhang, Senior Member, IEEE
TPAMI 2018
什么是shapelet?
shapelet是时间序列的子序列,具有判别性,其在某种意义上最大程度地代表类。
从树叶分类说起,自然是根据树叶的形状分类
将每个树叶用一个一维的时间序列来表示
根据shapelets分类
shapelets的优点
- 可解释性好
- 结果更精确,因为考虑的是局部特征,而非全局特征(很容易收到噪声和扭曲的影响)
- 速度更快,时间复杂度为O(mn)(m为时间序列长度,n为shapelet长度)
主要贡献
- USSL 无监督的shapelet学习,用于时间序列聚类,结合了pseudo-class labels, spectral analysis, regularized least-squares techniques and shapelet regularization
- 在UCR Time Series数据集上进行了验证,是state-of-the-art
目前方法存在的问题
-
最初通过扫描整个时间序列来获取shapelets的候选模板,时间复杂度很高
Synthetic Control dataset 包含600个时间序列, 长度为60,候选模板有1.098 × 10^6个
-
最近Grabocka等人提出一种基于回归学习的方法,寻找shapelets,大大减少了时间复杂度,但是需要标记
USSL框架
无监督的shapelet学习
-
Shapelet-transformed Representation
-
Pseudo-class Labels
-
Spectral Analysis
-
Least-squares Minimization
-
Shapelet Similarity Minimization
-
Unsupervised Salient Subsequence Learning
第一部分是谱正则化项,以保持时间序列之间的局部结构。第二部分通过最小化shapelets之间的相似性来使其多样化。第三部分则为学习最佳伪类标签和伪分类器的最小二乘正则化。
-
coordinate descent algorithm 坐标下降法
当要优化的函数含有多个自变量时,每次只更新一个自变量,固定其他自变量
实验
数据集:UCR Time Series 中的36个时间序列
评价指标
- Rand Index
- NMI(Normalized Mutual Information)
实验结果
-
等长的时间序列聚类
-
不等长的时间序列聚类
相关文章推荐
- 论文笔记之Learning Deep Representations for Graph Clustering
- DeepSense: a Unified Learning Framework for Time-Series Mobile Sensing Data Processing 论文笔记
- 论文笔记之---Learning Deep Feature Representations with Domain Guided Dropout for Person Re-identificatio
- 论文笔记之:Learning Cross-Modal Deep Representations for Robust Pedestrian Detection
- 多尺度R-CNN论文笔记(4): PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- 笔记-论文-PCANet A Simple Deep Learning Baseline for Image Classification?
- Joint Deep Learning For Pedestrian Detection(论文笔记-深度学习:行人检测)
- A Simple Pooling-Based Design for Real-Time Salient Object Detection------论文理解
- Anomaly Detection for Time Series Data with Deep Learning——本质分类正常和异常的行为,对于检测异常行为,采用预测正常行为方式来做
- 论文笔记:Perceptual Losses for Real-Time Style Transfer and Super-Resolution[doing]
- 论文笔记之:Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic
- 论文笔记——Deep Residual Learning for Image Recognition
- 论文笔记:PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
- Learning k for kNN Classification--论文笔记
- Deep Region and Multi-label Learning for Facial Action Unit Detection简要论文笔记
- FaceNet: A Unified Embedding for Face Recognition and Clustering论文笔记
- 【论文阅读笔记】RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechan
- Deep Learning for Content-Based Image Retrival:A Comprehensive Study 论文笔记
- 论文笔记 | Deep Residual Learning for Image Recognition
- 论文笔记之:Asynchronous Methods for Deep Reinforcement Learning