Time series classification (TSC) with deep learning has advanced markedly over the last decade, yet the mechanisms by which models identify and exploit discriminative patterns remain only partially understood, limiting interpretability. This paper surveys neurosymbolic approaches that aim either (i) to derive knowledge, by extracting human-readable rules and explanations from data, or (ii) to use knowledge, by injecting logical, physical, or structural constraints into learning. We review methods spanning Signal Temporal Logic (STL) formula extraction and temporal decision trees, neurosymbolic constraint synthesis and circuits, loss-based regularization including physics-informed objectives, probabilistic logic for event reasoning (e.g., DeepProbLog), and concept-level continual learning and distillation. We position these methods along common axes, explicit vs. implicit knowledge, applicability to TSC, capacity for knowledge discovery, prior knowledge requirements, and reproducibility, and discuss typical failure modes, including data bias and reasoning shortcuts. We also relate recent frugality-oriented designs that leverage non-trainable filters without sacrificing accuracy. Our analysis indicates that neurosymbolic techniques can provide actionable explanations and stronger inductive bias, with especially strong results on binary tasks, but face challenges in scaling to multi-class settings, handling shape-dependent phenomena, and ensuring transparent selection of learned rules. The survey offers a taxonomy and practical guidance for integrating symbolic structure with deep learning in TSC, highlighting open problems and promising directions.

Time series classification is widely used in many fields, but it often suffers from a lack of labeled data. To address this, researchers commonly apply data augmentation techniques that generate synthetic samples through transformations such as jittering, warping, or resampling. However, with an increasing number of available augmentation methods, it becomes difficult to choose the most suitable one for a given task. In many cases, this choice is based on intuition or visual inspection. Assessing the impact of this choice on classification accuracy requires training models, which is time-consuming and depends on the dataset. In this work, we adopt a generative model perspective and evaluate augmentation methods prior to training any classifier, using metrics that quantify both fidelity and diversity of the generated samples. We benchmark 22 augmentation techniques on 131 public datasets using eight metrics. Our results provide a practical and efficient way to compare augmentation methods without relying solely on classifier performance. The source code is publicly available: https://github.com/MSD-IRIMAS/Data-Augmentation-4-TSC

Time series machine learning (TSML) is a growing research field that spans a wide range of tasks. The popularity of established tasks such as classification, clustering, and extrinsic regression has, in part, been driven by the availability of benchmark datasets. An archive of 30 multivariate time series classification datasets, introduced in 2018 and commonly known as the UEA archive, has since become an essential resource cited in hundreds of publications. We present a substantial expansion of this archive that more than quadruples its size, from 30 to 133 classification problems. We also release preprocessed versions of datasets containing missing values or unequal length series, bringing the total number of datasets to 147. Reflecting the growth of the archive and the broader community, we rebrand it as the Multiverse archive to capture its diversity of domains. The Multiverse archive includes datasets from multiple sources, consolidating other collections and standalone datasets into a single, unified repository. Recognising that running experiments across the full archive is computationally demanding, we recommend a subset of the full archive called Multiverse-core (MV-core) for initial exploration. To support researchers in using the new archive, we provide detailed guidance and a baseline evaluation of established and recent classification algorithms, establishing performance benchmarks for future research. We have created a dedicated repository for the Multiverse archive that provides a common aeon and scikit-learn compatible framework for reproducibility, an extensive record of published results, and an interactive interface to explore the results.

Deep learning models for Time Series Classification (TSC) have achieved strong predictive performance but their high computational and memory requirements often limit deployment on resource-constrained devices. While structured pruning can address these issues by removing redundant filters, existing methods typically rely on manually tuned hyperparameters such as pruning ratios which limit scalability and generalization across datasets. In this work, we propose Dynamic Structured Pruning (DSP), a fully automatic, structured pruning framework for convolution-based TSC models. DSP introduces an instance-wise sparsity loss during training to induce channel-level sparsity, followed by a global activation analysis to identify and prune redundant filters without needing any predefined pruning ratio. This work tackles computational bottlenecks of deep TSC models for deployment on resource-constrained devices. We validate DSP on 128 UCR datasets using two different deep state-of-the-art architectures: LITETime and InceptionTime. Our approach achieves an average compression of 58

Global Average Pooling (GAP) has become a standard aggregation method in deep learning models for Time Series Classification (TSC), yet its effectiveness has recently been questioned by the research community. In this work, we conduct an extensive empirical investigation into the validity of GAP as an aggregation mechanism by comparing it to a diverse set of alternative methods. These include pooling-based, feature-based, and learnable aggregation techniques, evaluated across two well-established univariate (UCR) and multivariate (UEA) TSC benchmarks. Our results reveal that GAP remains highly competitive, consistently achieving strong classification performance with minimal computational overhead. Importantly, none of the alternative methods were able to statistically significantly outperform GAP, either in terms of accuracy or efficiency. Furthermore, we show that parametrized and complex aggregators, such as those based on Recurrent Neural Networks, often degrade performance, reinforcing the principle that simpler, non-parametric methods like GAP are not only sufficient but often preferable. This study reaffirms GAP as a robust and efficient choice for aggregation in deep neural networks for TSC tasks. All of our experimental results and source code are publicly available to ensure the reproducibility of our work and also to allow the community to use the raw results for further research.

The International StepUP Competition Series was launched to advance research in pressure-based footstep biometrics through a standardized and challenging evaluation framework. Using the large-scale StepUP-P150 dataset (with more than 200,000 high-resolution dynamic footsteps from 150 individuals) and a previously unreleased test set, the 2nd edition of the competition addressed three key challenges: (1) generalization to unseen users with limited enrollment data, (2) robustness to domain shift caused by variations in footwear and walking speed and (3) effective fusion of paired left-right footsteps. While the first two challenges built on the inaugural competition, this edition introduced more extreme cross-domain conditions and moved beyond isolated footsteps to stride-level verification, enabling new opportunities for representation learning and inter-step information fusion. The competition attracted 26 registrants from academia and industry, with a best equal error rate of 8.00

Automated assessment of human motion plays a vital role in rehabilitation, enabling objective evaluation of patient performance and progress. Unlike general human activity recognition, rehabilitation motion assessment focuses on analyzing the quality of movement within the same action class, requiring the detection of subtle deviations from ideal motion. Recent advances in deep learning and video-based skeleton extraction have opened new possibilities for accessible, scalable motion assessment using affordable devices such as smartphones or webcams. However, the field lacks standardized benchmarks, consistent evaluation protocols, and reproducible methodologies, limiting progress and comparability across studies. In this work, we address these gaps by (i) aggregating existing rehabilitation datasets into a unified archive called Rehab-Pile, (ii) proposing a general benchmarking framework for evaluating deep learning methods in this domain, and (iii) conducting extensive benchmarking of multiple architectures across classification and regression tasks. All datasets and implementations are released to the community to support transparency and reproducibility. This paper aims to establish a solid foundation for future research in automated rehabilitation assessment and foster the development of reliable, accessible, and personalized rehabilitation solutions. The datasets, source-code and results of this article are all publicly available.

Deep learning has achieved remarkable success in various domains including time series analysis, computer vision and natural language processing. However, high computational and memory demands of state-of-the-art architectures pose challenges for deployment in resource-limited environments. Knowledge Distillation (KD) addresses this by transferring knowledge from a large teacher model to a smaller, more efficient student model while maintaining competitive performance. In this work, we investigate the effectiveness of KD for Time Series Classification (TSC) across three architectures: the classical Fully Convolutional Network (FCN), the convolutional Inception model and the transformer-based ConvTran model. We evaluate our approach on UCR Archive, the largest benchmark repository of time series datasets, by modifying architectural components such as convolutional filters, Inception modules and attention heads across the three architectures. Our results consistently show that KD most effectively benefits student models of intermediate complexity across all three architectures, with the distilled FCN student reducing parameters by a factor of 38, the distilled Inception student achieving nearly the same performance as the teacher with 42

Despite advances in integrating reasoning based on belief functions to generalise probabilistic representations, distance-to-prototype-based evidential deep neural networks are still emerging and require further consolidation. Existing studies in segmentation or classification tasks typically perform prior initialisation and do not address or mitigate the potential conflicts that may arise during fusion. This work investigates high-conflict scenarios within an evidential neural network for segmentation in autonomous driving, focusing on the distance-to-prototypes component, where prototypes, derived from feature maps, serve as sources of evidence and may yield contradictory information. Conflict is mitigated through parameter adjustments within the evidential reasoning, enhancing consistency before fusion. This enables more reliable data integration and a valid application of fusion rules and decision-making processes. The proposed rectification is validated on two prototype configurations of a deep evidential lidar-camera cross-fusion architecture, using two distance-based decision strategies and adapted metrics. The impact on the network's predictions is demonstrated through qualitative and quantitative results on road detection tasks with the KITTI dataset.

The development of generative artificial intelligence for human motion generation has expanded rapidly, necessitating a unified evaluation framework. This paper presents a detailed review of eight evaluation metrics for human motion generation, highlighting their unique features and shortcomings. We propose standardized practices through a unified evaluation setup to facilitate consistent model comparisons. Additionally, we introduce a novel metric that assesses diversity in temporal distortion by analyzing warping diversity, thereby enhancing the evaluation of temporal data. We also conduct experimental analyses of three generative models using two publicly available datasets, offering insights into the interpretation of each metric in specific case scenarios. Our goal is to offer a clear, user-friendly evaluation framework for newcomers, complemented by publicly accessible code: https://github.com/MSD-IRIMAS/Evaluating-HMG.

Ensemble methods have played a crucial role in achieving state-of-the-art (SOTA) performance across various machine learning tasks by leveraging the diversity of features learned by individual models. In Time Series Classification (TSC), ensembles have proven highly effective whether based on neural networks (NNs) or traditional methods like HIVE-COTE. However most existing NN-based ensemble methods for TSC train multiple models with identical architectures and configurations. These ensembles aggregate predictions without explicitly promoting diversity which often leads to redundant feature representations and limits the benefits of ensembling. In this work, we introduce a diversity-driven ensemble learning framework that explicitly encourages feature diversity among neural network ensemble members. Our approach employs a decorrelated learning strategy using a feature orthogonality loss applied directly to the learned feature representations. This ensures that each model in the ensemble captures complementary rather than redundant information. We evaluate our framework on 128 datasets from the UCR archive and show that it achieves SOTA performance with fewer models. This makes our method both efficient and scalable compared to conventional NN-based ensemble approaches.

While human motion analysis has been widely addressed in recent years, the specific task of rehabilitation motion assessment remains challenging due to the lack of available annotated data. To overcome this challenge, data augmentation can be considered. However, classical augmentation techniques applied to human motion sequences often result in meaningless movements. Moreover, in rehabilitation assessment, labels are often continuous values illustrating the quality of a movement. Hence, associating a continuous label to augmented data is not straightforward. In this work, we propose to address data augmentation using an averaging method, called shapeDBA, adapted to rehabilitation motion sequences represented as multivariate time series. We extend the original proposal by weighting the average, hence allowing us to infer continuous labels associated to augmented motion sequences. We evaluated our proposed method on the Kimore dataset. Experimental results show that our method generates coherent rehabilitation sequences that can be efficiently used to extend a small dataset for rehabilitation assessment.

Deep learning models have been shown to be a powerful solution for Time Series Classification (TSC). State-of-the-art architectures, while producing promising results on the UCR and the UEA archives, present a high number of trainable parameters. This can lead to long training with high CO2 emission, power consumption and possible increase in the number of FLoating-point Operation Per Second (FLOPS). In this paper, we present a new architecture for TSC, the Light Inception with boosTing tEchnique (LITE) with only 2.34% of the number of parameters of the state-of-the-art InceptionTime model, while preserving performance. This architecture, with only 9, 814 trainable parameters due to the usage of DepthWise Separable Convolutions (DWSC), is boosted by three techniques: multiplexing, custom filters, and dilated convolution. The LITE architecture, trained on the UCR, is 2.78 times faster than InceptionTime and consumes 2.79 times less CO2 and power, while achieving an average accuracy of 84.62% compared to 84.91% with InceptionTime. To evaluate the performance of the proposed architecture on multivariate time series data, we adapt LITE to handle multivariate time series, we call this version LITEMV. To bring theory into application, we also conducted experiments using LITEMV on multivariate time series representing human rehabilitation movements, showing that LITEMV not only is the most efficient model but also the best performing for this application on the Kimore dataset, a skeleton-based human rehabilitation exercises dataset. Moreover, to address the interpretability of LITEMV, we present a study using Class Activation Maps to understand the classification decision taken by the model during evaluation.

MOOCs offer free and open access to a wide audience, but completion rates remain low, often due to a lack of personalized content. To address this issue, it is essential to predict learner performance in order to provide tailored feedback. Behavioral traces-such as clicks and events-can be analyzed as time series to anticipate learners' outcomes. This work compares multivariate time series classification methods to identify at-risk learners at different stages of the course (after 5, 10 weeks, etc.). The experimental evaluation, conducted on the Open University Learning Analytics Dataset (OULAD), focuses on three courses: two in STEM and one in SHS. Preliminary results show that the evaluated approaches are promising for predicting learner failure in MOOCs. The analysis also suggests that prediction accuracy is influenced by the amount of recorded interactions, highlighting the importance of rich and diverse behavioral data.

Human motion generation is an important research domain addressed by a significant amount of work in the recent years, with the availability of new datasets captured either from sensors or cameras. Most of the existing approaches are based on the Variational Auto-Encoder (VAE) architecture using Recurrent Neural Networks (RNN) or Transformers. In this paper, we propose to handle human motion sequences as Multivariate Time Series (MTS), and construct a VAE based on Convolutional Neural Networks (CNN). Furthermore, the proposed architecture uses an action classification task to add the conditioning aspect to the generative model. Our proposed Supervised VAE (SVAE) achieves competitive results on the HumanAct12 dataset, both in terms of quality and diversity of the generated sequences.

While automatic monitoring and coaching of exercises are showing encouraging results in non-medical applications, they still have limitations such as errors and limited use contexts. To allow the development and assessment of physical rehabilitation by an intelligent tutoring system, we identify in this article four challenges to address and propose a medical database of clinical patients carrying out low back-pain rehabilitation exercises. The dataset includes 3D Kinect skeleton positions and orientations, RGB videos, 2D skeleton data, and medical annotations to assess the correctness, and error classification and localisation of body part and timespan. Along this dataset, we perform a complete research path, from data collection to processing, and finally a small benchmark. We evaluated on the dataset two baseline movement recognition algorithms, pertaining to two different approaches: the probabilistic approach with a Gaussian Mixture Model (GMM), and the deep learning approach with a LongShort Term Memory (LSTM). This dataset is valuable because it includes rehabilitation relevant motions in a clinical setting with patients in their rehabilitation program, using a cost-effective, portable, and convenient sensor, and because it shows the potential for improvement on these challenges.