English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Time Series Classification (TSC) is essential in many fields, such asmedicine, environmental science and finance, enabling tasks like diseasediagnosis, anomaly detection, and stock price analysis. Machine learning modelsfor TSC like Recurrent Neural Networks and InceptionTime, while successful innumerous applications, can face scalability limitations due to intensivecomputational requirements. To address this, efficient models such as ROCKETand its derivatives have emerged, simplifying training and achievingstate-of-the-art performance by utilizing a large number of randomly generatedfeatures from time series data. However, due to their random nature, most ofthe generated features are redundant or non-informative, adding unnecessarycomputational load and compromising generalization. Here, we introduceSequential Feature Detachment (SFD) as a method to identify and prune thesenon-essential features. SFD uses model coefficients to estimate featureimportance and, unlike previous algorithms, can handle large feature setswithout the need for complex hyperparameter tuning. Testing on the UCR archivedemonstrates that SFD can produce models with $10\%$ of the original featureswhile improving the accuracy $0.2\%$ on the test set. We also present anend-to-end procedure for determining an optimal balance between the number offeatures and model accuracy, called Detach-ROCKET. When applied to the largestbinary UCR dataset, Detach-ROCKET is able to improve test accuracy by $0.6\%$while reducing the number of features by $98.9\%$. Thus, our proposed procedureis not only lightweight to train and effective in reducing model size andenhancing generalization, but its significant reduction in feature count alsopaves the way for feature interpretation.

Detach-ROCKET: Sequential feature selection for time series classification with random convolutional kernels