arXiv:2512.08246v1 Announce Type: new 
Abstract: Classical Time Series Classification algorithms are dominated by feature engineering strategies. One of the most prominent of these transforms is ROCKET, which achieves strong performance through random kernel features. We introduce SPROCKET (Selected Prototype Random Convolutional Kernel Transform), which implements a new feature engineering strategy based on prototypes. On a majority of the UCR and UEA Time Series Classification archives, SPROCKET achieves performance comparable to existing convolutional algorithms and the new MR-HY-SP ( MultiROCKET-HYDRA-SPROCKET) ensemble's average accuracy ranking exceeds HYDRA-MR, the previous best convolutional ensemble's performance. These experimental results demonstrate that prototype-based feature transformation can enhance both accuracy and robustness in time series classification.

تم تقديم SPROCKET، وهي استراتيجية جديدة لهندسة الميزات تعتمد على النماذج الأولية، لتحسين تصنيف السلاسل الزمنية، مما يوسع من قدرات خوارزمية ROCKET الحالية. تشير النتائج التجريبية إلى أن SPROCKET تحقق أداءً قابلاً للمقارنة مع الخوارزميات التلافيفية الرائدة عبر أرشيفات تصنيف السلاسل الزمنية UCR وUEA.

SPROCKET, una nueva estrategia de ingeniería de características basada en prototipos, ha sido introducida para mejorar la clasificación de series temporales, ampliando las capacidades del algoritmo ROCKET existente. Los resultados experimentales indican que SPROCKET logra un rendimiento comparable a los principales algoritmos convolucionales en los archivos de clasificación de series temporales UCR y UEA.

SPROCKET, une nouvelle stratégie d'ingénierie des caractéristiques basée sur des prototypes, a été introduite pour améliorer la classification des séries temporelles, étendant les capacités de l'algorithme ROCKET existant. Les résultats expérimentaux indiquent que SPROCKET atteint des performances comparables aux principaux algorithmes convolutionnels dans les archives de classification des séries temporelles UCR et UEA.

SPROCKET, a new feature engineering strategy based on prototypes, has been introduced to enhance time series classification, extending the capabilities of the existing ROCKET algorithm. Experimental results indicate that SPROCKET achieves performance comparable to leading convolutional algorithms across UCR and UEA Time Series Classification archives.

SPROCKET: Extending ROCKET to Distance-Based Time-Series Transformations With Prototypes

arXiv:2508.04503v2 Announce Type: replace 
Abstract: Multivariate time series classification supports applications from wearable sensing to biomedical monitoring and demands models that can capture both short-term patterns and longer-range temporal dependencies. Despite recent advances, Transformer and CNN models often remain computationally heavy and rely on many parameters. This work presents PRISM(Per-channel Resolution Informed Symmetric Module), a lightweight fully convolutional classifier. Operating in a channel-independent manner, in its early stage it applies a set of multi-resolution symmetric convolutional filters. This symmetry enforces structural constraints inspired by linear-phase FIR filters from classical signal processing, effectively halving the number of learnable parameters within the initial layers while preserving the full receptive field. Across the diverse UEA multivariate time-series archive as well as specific benchmarks in human activity recognition, sleep staging, and biomedical signals, PRISM matches or outperforms state-of-the-art CNN and Transformer models while using significantly fewer parameters and markedly lower computational cost. By bringing a principled signal processing prior into a modern neural architecture, PRISM offers an effective and computationally economical solution for multivariate time series classification.

تم تقديم PRISM كتصنيف خفيف الوزن للسلاسل الزمنية المتعددة المتغيرات، باستخدام طبقات تلافيفية متناظرة متعددة الدقة لالتقاط الأنماط قصيرة المدى والاعتماد الزمني طويل المدى بكفاءة. يقلل هذا النموذج بشكل كبير من عدد المعلمات القابلة للتعلم مع الحفاظ على الأداء عبر مجموعة متنوعة من المعايير، بما في ذلك التعرف على الأنشطة البشرية واكتشاف حالات النوم.

PRISM se ha presentado como un clasificador convolucional ligero para la clasificación de series temporales multivariadas, utilizando capas de convolución simétricas de multi-resolución para capturar de manera eficiente tanto patrones a corto plazo como dependencias temporales a más largo plazo. Este modelo reduce significativamente el número de parámetros aprendibles mientras mantiene el rendimiento en varios puntos de referencia, incluida la reconocimiento de actividades humanas y la detección de estados de sueño.

PRISM a été introduit comme un classificateur convolutif léger pour la classification des séries temporelles multivariées, utilisant des couches de convolution symétriques à multi-résolution pour capturer efficacement à la fois les motifs à court terme et les dépendances temporelles à plus long terme. Ce modèle réduit considérablement le nombre de paramètres apprenables tout en maintenant des performances sur divers benchmarks, y compris la reconnaissance d'activités humaines et la détection des états de sommeil.

PRISM has been introduced as a lightweight fully convolutional classifier for multivariate time series classification, utilizing symmetric multi-resolution convolutional layers to efficiently capture both short-term patterns and longer-range dependencies. This model significantly reduces the number of learnable parameters while maintaining performance across various benchmarks, including human activity recognition and sleep state detection.

PRISM: Lightweight Multivariate Time-Series Classification through Symmetric Multi-Resolution Convolutional Layers

arXiv:2512.06666v1 Announce Type: new 
Abstract: Time series classification faces a fundamental trade-off between accuracy and computational efficiency. While comprehensive ensembles like HIVE-COTE 2.0 achieve state-of-the-art accuracy, their 340-hour training time on the UCR benchmark renders them impractical for large-scale datasets. We investigate whether targeted combinations of two efficient algorithms from complementary paradigms can capture ensemble benefits while maintaining computational feasibility. Combining Hydra (competing convolutional kernels) and Quant (hierarchical interval quantiles) across six ensemble configurations, we evaluate performance on 10 large-scale MONSTER datasets (7,898 to 1,168,774 training instances). Our strongest configuration improves mean accuracy from 0.829 to 0.836, succeeding on 7 of 10 datasets. However, prediction-combination ensembles capture only 11% of theoretical oracle potential, revealing a substantial meta-learning optimization gap. Feature-concatenation approaches exceeded oracle bounds by learning novel decision boundaries, while prediction-level complementarity shows moderate correlation with ensemble gains. The central finding: the challenge has shifted from ensuring algorithms are different to learning how to combine them effectively. Current meta-learning strategies struggle to exploit the complementarity that oracle analysis confirms exists. Improved combination strategies could potentially double or triple ensemble gains across diverse time series classification applications.

تستكشف الدراسة التوازن بين الدقة والكفاءة الحاسوبية في تصنيف السلاسل الزمنية، مع تسليط الضوء على قيود التجميعات الشاملة مثل HIVE-COTE 2.0، التي تتطلب وقت تدريب طويل. من خلال دمج خوارزميات Hydra وQuant، تقيم الأبحاث الأداء عبر عشرة مجموعات بيانات MONSTER كبيرة الحجم، محققة تحسينًا في الدقة المتوسطة من 0.829 إلى 0.836 في سبع مجموعات بيانات. ومع ذلك، تكشف النتائج عن فجوة كبيرة في تحسين التعلم الميتا، حيث تلتقط مجموعات التنبؤ فقط 11% من الإمكانات النظرية.

El estudio explora el compromiso entre precisión y eficiencia computacional en la clasificación de series temporales, destacando las limitaciones de conjuntos completos como HIVE-COTE 2.0, que requieren un tiempo de entrenamiento extenso. Al combinar los algoritmos Hydra y Quant, la investigación evalúa el rendimiento en diez conjuntos de datos MONSTER a gran escala, logrando una mejora en la precisión media de 0.829 a 0.836 en siete conjuntos de datos. Sin embargo, los hallazgos revelan una brecha significativa en la optimización del meta-aprendizaje, con los conjuntos de combinaciones de pre…

L'étude examine le compromis entre précision et efficacité computationnelle dans la classification des séries temporelles, mettant en évidence les limites des ensembles complets comme HIVE-COTE 2.0, qui nécessitent un temps d'entraînement considérable. En combinant les algorithmes Hydra et Quant, la recherche évalue les performances sur dix ensembles de données MONSTER à grande échelle, atteignant une amélioration de la précision moyenne de 0,829 à 0,836 dans sept ensembles de données. Cependant, les résultats révèlent un écart d'optimisation en méta-apprentissage significatif, les ensembles d…

The study explores the trade-off between accuracy and computational efficiency in time series classification, highlighting the limitations of comprehensive ensembles like HIVE-COTE 2.0, which require extensive training time. By combining Hydra and Quant algorithms, the research evaluates performance across ten large-scale MONSTER datasets, achieving a mean accuracy improvement from 0.829 to 0.836 in seven datasets. However, the findings reveal a significant meta-learning optimization gap, with prediction-combination ensembles capturing only 11% of theoretical potential.

SPROCKET: Extending ROCKET to Distance-Based Time-Series Transformations With Prototypes

Was this article worth reading? Share it

B2B Rocket

RandomGenerator.AI

Random Generator Pro