ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels

Dempster, Angus; Petitjean, François; Webb, Geoffrey I.

doi:10.1007/s10618-020-00701-z

Cited by 611 publications

(414 citation statements)

References 34 publications

Supporting

Mentioning

409

Contrasting

Unclassified

Order By: Relevance

“…However, the real winner of this experimental analysis is ROCKET. We did not originally include ROCKET in this study (an early iteration of this paper is available on ArXiv (Pasos-Ruiz et al 2020)) because multivariate capability is listed as future work in the related publication (Dempster et al 2020). However, the authors of ROCKET contributed their code to the sktime toolkit with multivariate functionality, so we could include it in the study.…”

Section: Discussionmentioning

confidence: 99%

“…Nevertheless, some advice for practitioners for a starting point in an analysis is always helpful. We conclude that one recently published algorithm, ROCKET (Dempster et al 2020), is our recommended choice due to high overall accuracy and remarkably fast training time.…”

Section: Introductionmentioning

confidence: 92%

“…The Random Convolutional Kernel Transform, ROCKET (Dempster et al 2020) uses a large number of random convolution kernels in conjunction with a linear classifier (ridge regression or logistic regression). Every kernel is applied to each instance.…”

Section: The Random Convolutional Kernel Transform (Rocket)mentioning

confidence: 99%

See 2 more Smart Citations

The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances

Ruiz

Flynn

Large

et al. 2020

Data Min Knowl Disc

311

156

View full text Add to dashboard Cite

Time Series Classification (TSC) involves building predictive models for a discrete target variable from ordered, real valued, attributes. Over recent years, a new set of TSC algorithms have been developed which have made significant improvement over the previous state of the art. The main focus has been on univariate TSC, i.e. the problem where each case has a single series and a class label. In reality, it is more common to encounter multivariate TSC (MTSC) problems where the time series for a single case has multiple dimensions. Despite this, much less consideration has been given to MTSC than the univariate case. The UCR archive has provided a valuable resource for univariate TSC, and the lack of a standard set of test problems may explain why there has been less focus on MTSC. The UEA archive of 30 MTSC problems released in 2018 has made comparison of algorithms easier. We review recently proposed bespoke MTSC algorithms based on deep learning, shapelets and bag of words approaches. If an algorithm cannot naturally handle multivariate data, the simplest approach to adapt a univariate classifier to MTSC is to ensemble it over the multivariate dimensions. We compare the bespoke algorithms to these dimension independent approaches on the 26 of the 30 MTSC archive problems where the data are all of equal length. We demonstrate that four classifiers are significantly more accurate than the benchmark dynamic time warping algorithm and that one of these recently proposed classifiers, ROCKET, achieves significant improvement on the archive datasets in at least an order of magnitude less time than the other three.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 92%

See 1 more Smart Citation

The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances

Ruiz

Flynn

Large

et al. 2020

Data Min Knowl Disc

311

156

View full text Add to dashboard Cite

show abstract

“…The motivation for using the two aforementioned DSs is to establish whether accuracy of expertise classification is improved when fNIRS data is used in tandem with predicted emotion scores data (DS2) in comparison to when only fNIRS data is used (DS1). In addition, the classification prowess of the supervised classifiers are also gauged when these classifiers are given statistical features in comparison to random convolutional kernel transform (ROCKET) [ 38 ] features. A flowchart depicting the overall classification paradigm is presented in Figure 5 .…”

Section: Methodsmentioning

confidence: 99%

Single-Trial Recognition of Video Gamer’s Expertise from Brain Haemodynamic and Facial Emotion Responses

et al. 2021

View full text Add to dashboard Cite

With an increase in consumer demand of video gaming entertainment, the game industry is exploring novel ways of game interaction such as providing direct interfaces between the game and the gamers’ cognitive or affective responses. In this work, gamer’s brain activity has been imaged using functional near infrared spectroscopy (fNIRS) whilst they watch video of a video game (League of Legends) they play. A video of the face of the participants is also recorded for each of a total of 15 trials where a trial is defined as watching a gameplay video. From the data collected, i.e., gamer’s fNIRS data in combination with emotional state estimation from gamer’s facial expressions, the expertise level of the gamers has been decoded per trial in a multi-modal framework comprising of unsupervised deep feature learning and classification by state-of-the-art models. The best tri-class classification accuracy is obtained using a cascade of random convolutional kernel transform (ROCKET) feature extraction method and deep classifier at 91.44%. This is the first work that aims at decoding expertise level of gamers using non-restrictive and portable technologies for brain imaging, and emotional state recognition derived from gamers’ facial expressions. This work has profound implications for novel designs of future human interactions with video games and brain-controlled games.

show abstract

“…However, every input instance will be considered to adjust the inner concept, which requires potentially a large buffer space and will bring a huge computation cost. Recent Deep Neural Network (DNN) approaches [24][25][26][27] on TSC are capable of tuning the model incrementally, but stay always in an awkward position for the lack of explainability, which is required by domains like healthcare where questions of accountability and transparency are particularly important.…”

Section: Introductionmentioning

confidence: 99%

Incremental and Adaptive Feature Exploration over Time Series Stream

Zuo

Zeitouni

Taher

2019

2019 IEEE International Conference on Big Data (Big Data)

View full text Add to dashboard Cite

Over past years, various attempts have been made at analysing Time Series (TS) which has been raising great interest of Data Mining community due to its special data format and broad application scenarios. An important aspect in TS analysis is Time Series Classification (TSC), which has been applied in medical diagnosis, human activity recognition, industrial troubleshooting, etc. Typically, all TSC work trains a stable model from an off-line TS dataset, without considering potential Concept Drift in streaming context. Domains like healthcare look to enrich the database gradually with more medical cases, or in astronomy, with human's growing knowledge about the universe, the theoretical basis for labelling data will change. The techniques applied in a stable TS dataset are then not adaptable in such dynamic scenarios (i.e. streaming context). Classical data stream analysis are biased towards vector or row data, where each attribute is independent to train an adaptive learning model, but rarely considers Time Series as a stream instance. Processing such type of data requires combining techniques in both communities of Time Series (TS) and Data Streams. To this end, by adopting the concepts of Shapelet and Matrix Profile, we conduct the first attempt to extract the adaptive features from Time Series Stream based on the Test-then-Train strategy, which is applicable in both contexts: a) under stable concept, learning model will be updated incrementally; b) for data source with Concept Drift, previous concepts that do not represent the current stream behavior will be discarded from the model.

show abstract

ROCKET: exceptionally fast and accurate time series classification using random convolutional kernels

Cited by 611 publications

References 34 publications

The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances

The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances

Single-Trial Recognition of Video Gamer’s Expertise from Brain Haemodynamic and Facial Emotion Responses

Incremental and Adaptive Feature Exploration over Time Series Stream

Contact Info

Product

Resources

About