Deep learning for time series classification: a review

Abstract

Time Series Classification (TSC) is an important and challenging problem in data mining. With the increase of time series data availability, hundreds of TSC algorithms have been proposed. Among these methods, only a few have considered Deep Neural Networks (DNNs) to perform this task. This is surprising as deep learning has seen very successful applications in the last years. DNNs have indeed revolutionized the field of computer vision especially with the advent of novel deeper architectures such as Residual and Convolutional Neural Networks. Apart from images, sequential data such as text and audio can also be processed with DNNs to reach state-of-the-art performance for document classification and speech recognition. In this article, we study the current state-ofthe-art performance of deep learning algorithms for TSC by presenting an empirical study of the most recent DNN architectures for TSC. We give an overview of the most successful deep learning applications in various time series domains under a unified taxonomy of DNNs for TSC. We also provide an open source deep learning framework to the TSC community where we implemented each of the compared approaches and evaluated them on a univariate TSC benchmark (the UCR/UEA archive) and 12 multivariate time series datasets. By training 8,730 deep learning models on 97 time series datasets, we propose the most exhaustive study of DNNs for TSC to date.

Estimation of neural network model parameters from local field potentials (LFPs)

“…The choice of using convolutional neural networks (CNNs) within the Keras framework (Chollet et al, 2015) for doing the parameter estimation was made out of convenience. Other machine learning techniques, see Ismail Fawaz et al (2018) for a recent review, could likely have done as good, or even better. Further, the architecture of the CNNs was not optimised in any systematic way.…”

Temporal Convolutional Neural Network for the Classification of Satellite Image Time Series

“…Two types of pooling have received most of the attention in the literature about image analysis: 1) the local max-pooling [72], and 2) the global average pooling [73]. For time series, global average pooling seems to have been more successful [9,37]. We want to see here if these previous results can be generalized for time series classification.…”

“…Table 4 displays the OA values as a function of reach for TempCNN. We study five size of filters f = {3, 5,9,17, 33} corresponding to a reach of 2, 4, 8, 16, and 32 days, respectively. Table 4 shows the maximum OA is reached for a reach of 8 days, with a similiar OA for 4 days.…”

“…They also question the choice of the classification algorithms: although traditional algorithms showed good performance for SITS classification [3], they do not explicitly use the temporal relationship between acquisitions. Concurrently, deep learning methods are appealing for this task [9]. Their ability to automatically extract useful representations has been demonstrated in machine learning [10], as well as on various remote sensing tasks [11].…”

Artificial neural network subgrid models of 2D compressible magnetohydrodynamic turbulence

“…The basic units of calculation in ANNs are called neurons, which are connected via weighted inputs that resemble synapses. These biologically inspired models have the proven capability of learning from data, which has accelerated the data-driven discovery revolution over the last decade [25][26][27][28][29].…”