A Complexity-Invariant Distance Measure for Time Series

By Gustavo E. A. P. A. Batista, Xiaoyue Wang, Eamonn J. Keogh. SDM, 2011

345 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Economics and Econometrics, Signal Processing

View Paper: https://www.semanticscholar.org/paper/2e310a585d0010d8ac90d35539aab72fd83b5a88

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Rohit J. Kate
2. Anthony J. Bagnall
3. Jason Lines
4. Aaron Bostrom
5. Jon Hills
6. Nick S. Jones
7. James Mapp
8. Johan J. Lukkien

In Economics and Econometrics

1. Rohit J. Kate
2. Anthony J. Bagnall
3. Jason Lines
4. Aaron Bostrom
5. Jon Hills
6. Nick S. Jones
7. James Mapp
8. Johan J. Lukkien

In Signal Processing

1. Rohit J. Kate
2. Anthony J. Bagnall
3. Jason Lines
4. Aaron Bostrom
5. Jon Hills
6. Ben D. Fulcher
7. Nick S. Jones
8. James Mapp

Citations in the corpus

• Multi-task Self-Supervised Learning for Human Activity Detection (2019, methodology): “We show that when training a CNN to predict generally known (time-series) transformations [7, 67] as a surrogate task, the model can learn features that are on a par with a fully-supervised network and far beŠer than unsupervised pre-training with an autoencoder.”
• Highly Comparative Feature-Based Time-Series Classification (2014, background): “For example, Batista et al. [46] used a simple new feature claimed to resemble ‘complexity’, to rescale conventional Euclidean distances calculated between time series, demonstrating an improvement in classification accuracy.”
• Time Series Classification Using Multi-Channels Deep Convolutional Neural Networks (2014, methodology): “Among these classification methods, the distance-based method k-Nearest Neighbor (k-NN) classification has been empirically proven to be very difficult to beat [4, 5].”
• TSclust: An R Package for Time Series Clustering (2014, methodology): “…by Batista et al. is very simple and consists in computing The CID method is intuitive, parameter-free, invariant to the complexity of time series, computationally efficient, and it has produced improvements in accuracy in several clustering experiments carried out in Batista et al. (2011).”
• Exploiting multi-channels deep convolutional neural networks for multivariate time series classification (2015, neutral): “On the other hand, more evidences have also shown that the Dynamic Time Warping (DTW) is the best sequence distance measurement for time series in many domains [2, 8–10].”
• Classification of time series by shapelet transformation (2013, background): “For example, Batista et al. state that “there is a plethora of classification algorithms that can be applied to time series; however, all of the current empirical evidence suggests that simple nearest neighbor classification is very difficult to beat” Batista et al. (2011).”
• Time-Series Classification with COTE: The Collective of Transformation-Based Ensembles (2015, background): “The establishment of the UCR repository for TSC problems [1] has engendered growth in the number of algorithms proposed for TSC (for example, see [2], [3], [4], [5], [6], [7], [8], [9]).”
• Using dynamic time warping distances as features for improved time series classification (2016, methodology): “Simply showing that a new classifier is better on multiple datasets and hoping that a new dataset will be one of them is falling for a subtle version of the Texas sharpshooter fallacy (Batista et al. 2011).”