Deep Learning for Time Series Classification: A Review

By H. I. Fawaz et al

Read the original document by opening this link in a new tab.

Table of Contents

1. Abstract
2. Introduction
3. Background
4. Main Contributions
5. Time Series Classification
6. Deep Learning for Time Series Classification
7. Multi Layer Perceptrons
8. Convolutional Neural Networks
9. Echo State Network
10. Conclusion

Summary

This document is a review article on Time Series Classification (TSC) focusing on the application of Deep Neural Networks (DNNs) for this task. It discusses the challenges in TSC, the historical methods used, and the recent advancements with DNNs. The paper presents an empirical study of various DNN architectures for TSC, evaluating their performance on univariate and multivariate time series datasets. It also introduces an open-source deep learning framework for TSC and explores methods to mitigate the black-box effect of DNNs. The main contributions include explaining how deep learning can be adapted to one-dimensional time series data, proposing a taxonomy of DNN applications for TSC, detailing nine deep learning models designed for TSC, evaluating these models on benchmark datasets, providing the community with an open-source framework, and investigating interpretability of DNN decisions.