Most methods for time series classification that attain state-of-the-art accuracy have high computational complexity, requiring significant training time even for smaller datasets, and are intractable for larger datasets. Additionally, many existing methods focus on a single type of feature such as shape or frequency. Building on the recent success of convolutional neural networks for time series classification, we show that simple linear classifiers using random convolutional kernels achieve state-of-the-art accuracy with a fraction of the computational expense of existing methods.
In the time series classification domain, shapelets are small time series that are discriminative for a certain class. It has been shown that classifiers are able to achieve state-of-the-art results on a plethora of datasets by taking as input distances from the input time series to different discriminative shapelets. Additionally, these shapelets can easily be visualized and thus possess an interpretable characteristic, making them very appealing in critical domains, such as the health care domain, where longitudinal data is ubiquitous. In this study, a new paradigm for shapelet discovery is proposed, which is based upon evolutionary computation. The advantages of the proposed approach are that (i) it is gradient-free, which could allow to escape from local optima more easily and to find suited candidates more easily and supports non-differentiable objectives, (ii) no brute-force search is required, which drastically reduces the computational complexity by several orders of magnitude, (iii) the total amount of shapelets and length of each of these shapelets are evolved jointly with the shapelets themselves, alleviating the need to specify this beforehand, (iv) entire sets are evaluated at once as opposed to single shapelets, which results in smaller final sets with less similar shapelets that result in similar predictive performances, and (v) discovered shapelets do not need to be a subsequence of the input time series. We present the results of experiments which validate the enumerated advantages.
The amount of available Earth observation data has increased dramatically in the recent years. Efficiently making use of the entire body information is a current challenge in remote sensing and demands for light-weight problem-agnostic models that do not require region- or problem-specific expert knowledge. End-to-end trained deep learning models can make use of raw sensory data by learning feature extraction and classification in one step solely from data. Still, many methods proposed in remote sensing research require implicit feature extraction through data preprocessing or explicit design of features. In this work, we compare recent deep learning models on crop type classification on raw and preprocessed Sentinel 2 data. We concentrate on the common neural network architectures for time series, i.e., 1D-convolutions, recurrence, a shallow random forest baseline, and focus on the novel self-attention architecture. Our central findings are that data preprocessing still increased the overall classification performance for all models while the choice of model was less crucial. Self-attention and recurrent neural networks, by their architecture, outperformed convolutional neural networks on raw satellite time series. We explore this by a feature importance analysis based on gradient back-propagation that exploits the differentiable nature of deep learning models. Further, we qualitatively show how self-attention scores focus selectively on few classification-relevant observations.
Time Series Classification (TSC) has been an important and challenging task in data mining, especially on multivariate time series and multi-view time series data sets. Meanwhile, transfer learning has been widely applied in computer vision and natural language processing applications to improve deep neural network's generalization capabilities. However, very few previous works applied transfer learning framework to time series mining problems. Particularly, the technique of measuring similarities between source domain and target domain based on dynamic representation such as density estimation with importance sampling has never been combined with transfer learning framework. In this paper, we first proposed a general adaptive transfer learning framework for multi-view time series data, which shows strong ability in storing inter-view importance value in the process of knowledge transfer. Next, we represented inter-view importance through some time series similarity measurements and approximated the posterior distribution in latent space for the importance sampling via density estimation techniques. We then computed the matrix norm of sampled importance value, which controls the degree of knowledge transfer in pre-training process. We further evaluated our work, applied it to many other time series classification tasks, and observed that our architecture maintained desirable generalization ability. Finally, we concluded that our framework could be adapted with deep learning techniques to receive significant model performance improvements.
Research into time series classification has tended to focus on the case of series of uniform length. However, it is common for real-world time series data to have unequal lengths. Differing time series lengths may arise from a number of fundamentally different mechanisms. In this work, we identify and evaluate two classes of such mechanisms -- variations in sampling rate relative to the relevant signal and variations between the start and end points of one time series relative to one another. We investigate how time series generated by each of these classes of mechanism are best addressed for time series classification. We perform extensive experiments and provide practical recommendations on how variations in length should be handled in time series classification.
Despite the eminent successes of deep neural networks, many architectures are often hard to transfer to irregularly-sampled and asynchronous time series that commonly occur in real-world datasets, especially in healthcare applications. This paper proposes a novel approach for classifying irregularly-sampled time series with unaligned measurements, focusing on high scalability and data efficiency. Our method SeFT (Set Functions for Time Series) is based on recent advances in differentiable set function learning, extremely parallelizable with a beneficial memory footprint, thus scaling well to large datasets of long time series and online monitoring scenarios. Furthermore, our approach permits quantifying per-observation contributions to the classification outcome. We extensively compare our method with existing algorithms on multiple healthcare time series datasets and demonstrate that it performs competitively whilst significantly reducing runtime.
Time series are series of values ordered by time. This kind of data can be found in many real world settings. Classifying time series is a difficult task and an active area of research. This paper investigates the use of transfer learning in Deep Neural Networks and a 2D representation of time series known as Recurrence Plots. In order to utilize the research done in the area of image classification, where Deep Neural Networks have achieved very good results, we use a Residual Neural Networks architecture known as ResNet. As preprocessing of time series is a major part of every time series classification pipeline, the method proposed simplifies this step and requires only few parameters. For the first time we propose a method for multi time series classification: Training a single network to classify all datasets in the archive with one network. We are among the first to evaluate the method on the latest 2018 release of the UCR archive, a well established time series classification benchmarking dataset.
We present sktime -- a new scikit-learn compatible Python library with a unified interface for machine learning with time series. Time series data gives rise to various distinct but closely related learning tasks, such as forecasting and time series classification, many of which can be solved by reducing them to related simpler tasks. We discuss the main rationale for creating a unified interface, including reduction, as well as the design of sktime's core API, supported by a clear overview of common time series tasks and reduction approaches.
In this paper, we present a new deep learning architecture for addressing the problem of supervised learning with sparse and irregularly sampled multivariate time series. The architecture is based on the use of a semi-parametric interpolation network followed by the application of a prediction network. The interpolation network allows for information to be shared across multiple dimensions of a multivariate time series during the interpolation stage, while any standard deep learning model can be used for the prediction network. This work is motivated by the analysis of physiological time series data in electronic health records, which are sparse, irregularly sampled, and multivariate. We investigate the performance of this architecture on both classification and regression tasks, showing that our approach outperforms a range of baseline and recently proposed models.
Deep neural networks (DNNs) have achieved state-of-the-art results on time series classification (TSC) tasks. In this work, we focus on leveraging DNNs in the often-encountered practical scenario where access to labeled training data is difficult, and where DNNs would be prone to overfitting. We leverage recent advancements in gradient-based meta-learning, and propose an approach to train a residual neural network with convolutional layers as a meta-learning agent for few-shot TSC. The network is trained on a diverse set of few-shot tasks sampled from various domains (e.g. healthcare, activity recognition, etc.) such that it can solve a target task from another domain using only a small number of training samples from the target task. Most existing meta-learning approaches are limited in practice as they assume a fixed number of target classes across tasks. We overcome this limitation in order to train a common agent across domains with each domain having different number of target classes, we utilize a triplet-loss based learning procedure that does not require any constraints to be enforced on the number of classes for the few-shot TSC tasks. To the best of our knowledge, we are the first to use meta-learning based pre-training for TSC. Our approach sets a new benchmark for few-shot TSC, outperforming several strong baselines on few-shot tasks sampled from 41 datasets in UCR TSC Archive. We observe that pre-training under the meta-learning paradigm allows the network to quickly adapt to new unseen tasks with small number of labeled instances.
sktime is an open source, Python based, sklearn compatible toolkit for time series analysis developed by researchers at the University of East Anglia (UEA), University College London and the Alan Turing Institute. A key initial goal for sktime was to provide time series classification functionality equivalent to that available in a related java package, tsml, also developed at UEA. We describe the implementation of six such classifiers in sktime and compare them to their tsml equivalents. We demonstrate correctness through equivalence of accuracy on a range of standard test problems and compare the build time of the different implementations. We find that there is significant difference in accuracy on only one of the six algorithms we look at (Proximity Forest). This difference is causing us some pain in debugging. We found a much wider range of difference in efficiency. Again, this was not unexpected, but it does highlight ways both toolkits could be improved.
The Large Synoptic Survey Telescope will complete its survey in 2022 and produce terabytes of imaging data each night. To work with this massive onset of data, automated algorithms to classify astronomical light curves are crucial. Here, we present a method for automated classification of photometric light curves for a range of astronomical objects. Our approach is based on the gradient boosting of decision trees, feature extraction and selection, and augmentation. The solution was developed in the context of The Photometric LSST Astronomical Time Series Classification Challenge (PLAsTiCC) and achieved one of the top results in the challenge.
This paper brings deep learning at the forefront of research into Time Series Classification (TSC). TSC is the area of machine learning tasked with the categorization (or labelling) of time series. The last few decades of work in this area have led to significant progress in the accuracy of classifiers, with the state of the art now represented by the HIVE-COTE algorithm. While extremely accurate, HIVE-COTE cannot be applied to many real-world datasets because of its high training time complexity in O(N2 * T4) for a dataset with N time series of length T. For example, it takes HIVE-COTE more than 8 days to learn from a small dataset with N = 1500 time series of short length T = 46. Meanwhile deep learning has received enormous attention because of its high accuracy and scalability. Recent approaches to deep learning for TSC have been scalable, but less accurate than HIVE-COTE. We introduce InceptionTime - an ensemble of deep Convolutional Neural Network (CNN) models, inspired by the Inception-v4 architecture. Our experiments show that InceptionTime is on par with HIVE-COTE in terms of accuracy while being much more scalable: not only can it learn from 1,500 time series in one hour but it can also learn from 8M time series in 13 hours, a quantity of data that is fully out of reach of HIVE-COTE.
We use standard deep neural networks to classify univariate time series generated by discrete and continuous dynamical systems based on their chaotic or non-chaotic behaviour. Our approach to circumvent the lack of precise models for some of the most challenging real-life applications is to train different neural networks on a data set from a dynamical system with a basic or low-dimensional phase space and then use these networks to classify univariate time series of a dynamical system with more intricate or high-dimensional phase space. We illustrate this generalisation approach using the logistic map, the sine-circle map, the Lorenz system, and the Kuramoto--Sivashinsky equation. We observe that a convolutional neural network without batch normalization layers outperforms state-of-the-art neural networks for time series classification and is able to generalise and classify time series as chaotic or not with high accuracy.
Early time series classification (eTSC) is the problem of classifying a time series after as few measurements as possible with the highest possible accuracy. The most critical issue of any eTSC method is to decide when enough data of a time series has been seen to take a decision: Waiting for more data points usually makes the classification problem easier but delays the time in which a classification is made; in contrast, earlier classification has to cope with less input data, often leading to inferior accuracy. The state-of-the-art eTSC methods compute a fixed optimal decision time assuming that every times series has the same defined start time (like turning on a machine). However, in many real-life applications measurements start at arbitrary times (like measuring heartbeats of a patient), implying that the best time for taking a decision varies heavily between time series. We present TEASER, a novel algorithm that models eTSC as a two two-tier classification problem: In the first tier, a classifier periodically assesses the incoming time series to compute class probabilities. However, these class probabilities are only used as output label if a second-tier classifier decides that the predicted label is reliable enough, which can happen after a different number of measurements. In an evaluation using 45 benchmark datasets, TEASER is two to three times earlier at predictions than its competitors while reaching the same or an even higher classification accuracy. We further show TEASER's superior performance using real-life use cases, namely energy monitoring, and gait detection.
Dictionary based classifiers are a family of algorithms for time series classification (TSC), that focus on capturing the frequency of pattern occurrences in a time series. The ensemble based Bag of Symbolic Fourier Approximation Symbols (BOSS) was found to be a top performing TSC algorithm in a recent evaluation, as well as the best performing dictionary based classifier. A recent addition to the category, the Word Extraction for Time Series Classification (WEASEL), claims an improvement on this performance. Both of these algorithms however have non-trivial scalability issues, taking a considerable amount of build time and space on larger datasets. We evaluate changes to the way BOSS chooses classifiers for its ensemble, replacing its parameter search with random selection. This change allows for the easy implementation of contracting, setting a build time limit for the classifier and check-pointing, saving progress during the classifiers build. To differentiate between the two BOSS ensemble methods we refer to our randomised version as RBOSS. Additionally we test the application of common ensembling techniques to help retain accuracy from the loss of the BOSS parameter search. We achieve a significant reduction in build time without a significant change in accuracy on average when compared to BOSS by creating a size $n$ weighted ensemble selecting the best performers from $k$ randomly chosen parameter sets. Our experiments are conducted on datasets from the recently expanded UCR time series archive. We demonstrate the usability improvements to RBOSS with a case study using a large whale acoustics dataset for which BOSS proved infeasible.
The art of systematic financial trading evolved with an array of approaches, ranging from simple strategies to complex algorithms all relying, primary, on aspects of time-series analysis. Recently, after visiting the trading floor of a leading financial institution, we noticed that traders always execute their trade orders while observing images of financial time-series on their screens. In this work, we built upon the success in image recognition and examine the value in transforming the traditional time-series analysis to that of image classification. We create a large sample of financial time-series images encoded as candlestick (Box and Whisker) charts and label the samples following three algebraically-defined binary trade strategies. Using the images, we train over a dozen machine-learning classification models and find that the algorithms are very efficient in recovering the complicated, multiscale label-generating rules when the data is represented visually. We suggest that the transformation of continuous numeric time-series classification problem to a vision problem is useful for recovering signals typical of technical analysis.
Time series with non-uniform intervals occur in many applications, and are difficult to model using standard recurrent neural networks (RNNs). We generalize RNNs to have continuous-time hidden dynamics defined by ordinary differential equations (ODEs), a model we call ODE-RNNs. Furthermore, we use ODE-RNNs to replace the recognition network of the recently-proposed Latent ODE model. Both ODE-RNNs and Latent ODEs can naturally handle arbitrary time gaps between observations, and can explicitly model the probability of observation times using Poisson processes. We show experimentally that these ODE-based models outperform their RNN-based counterparts on irregularly-sampled data.
Modern well-performing approaches to neural decoding are based on machine learning models such as decision tree ensembles and deep neural networks. The wide range of algorithms that can be utilized to learn from neural spike trains, which are essentially time-series data, results in the need for diverse and challenging benchmarks for neural decoding, similar to the ones in the fields of computer vision and natural language processing. In this work, we propose a spike train classification benchmark, based on open-access neural activity datasets and consisting of several learning tasks such as stimulus type classification, animal's behavioral state prediction, and neuron type identification. We demonstrate that an approach based on hand-crafted time-series feature engineering establishes a strong baseline performing on par with state-of-the-art deep learning-based models for neural decoding. We release the code allowing to reproduce the reported results.
In this paper, a 1d convolutional neural network is designed for classification tasks of plant leaves. This network based classifier is analyzed in two directions. In the forward direction, the proposed network can be used in two ways: a classifier and an automatic feature extractor. As a classifier, it takes the simple centroid contour distance curve as the single feature and achieves comparable performance with state-of-art methods that usually require multiple extracted features. As a feature extractor, it produces nearly linear separable features, hence can be used together with other classifiers such as support vector machines to provide better performance. The proposed network adopts simple 1d input and is generally applicable for other tasks such as classifying one dimensional time series in an end-to-end fashion without changes. Experiments on some benchmark datasets show this architecture can provide classification accuracies that are comparable or higher than some existing methods. In the backward direction, methods like gradient-weighted class activation mapping and maximum activation map of neurons in the classification layer with respect to inputs are performed to help investigate and further validate that hidden signatures helping trigger the trained classifier's specific decisions can be human interpretable. Code for the paper is available at https://github.com/dykuang/Leaf_Project.