mlsanityData to 7 Oct 2026

MSTN: A Lightweight and Fast Model for General TimeSeries Analysis

arXiv 2511.20577 ↗First posted 25 Nov 2025cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Real-world time series often exhibit strong non-stationarity, complex nonlinear dynamics, and behavior expressed across multiple temporal scales, from rapid local fluctuations to slow-evolving long-range trends. However, many contemporary architectures impose rigid, fixed-scale structural priors such as patch-based tokenization, predefined receptive fields, or frozen backbone encoders - which can over-regularize temporal dynamics and limit adaptability to abrupt high-magnitude events. To handle this, we introduce the Multi-scale Temporal Network (MSTN), a hybrid neural architecture grounded in an Early Temporal Aggregation principle. MSTN integrates three complementary components: (i) a multi-scale convolutional encoder that captures fine-grained local structure; (ii) a sequence modeling module that learns long-range dependencies through either recurrent or attention-based mechanisms; and (iii) a self-gated fusion stage incorporating squeeze-excitation and a single dense layer to dynamically reweight and fuse multi-scale representations. ETA ensures downstream modules operate in O(1) time, while the encoder retains O(L^2) (Transformer) or O(L) (BiLSTM). This design enables MSTN to flexibly model temporal patterns spanning milliseconds to extended horizons, while avoiding the computational burden typically associated with long-context models. Across extensive benchmarks covering imputation, long-term forecasting, classification, and cross-dataset generalization, MSTN achieves state-of-the-art performance, establishing new best results on 21 of 27 datasets, while remaining lightweight (~0.40M params for MSTN-BiLSTM and ~1.06M for MSTN-Transformer) and suitable for low-latency inference (<1 sec, often in milliseconds), resource-constrained deployment.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 30 of its 59 references that are in the map. First posted on arXiv 25 Nov 2025.

Builds on

30 of 59 references.

References · 30

  1. 2022A Time Series is Worth 64 Words: Long-term Forecasting with Transformerscited by 1,981
  2. 2021Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecastingcited by 1,654
  3. 2018An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modelingcited by 1,447
  4. 2022Are Transformers Effective for Time Series Forecasting?cited by 1,400
  5. 2022FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecastingcited by 1,200
  6. 2023iTransformer: Inverted Transformers Are Effective for Time Series Forecastingcited by 906
  7. 2017Modeling Long- and Short-Term Temporal Patterns with Deep Neural Networkscited by 869
  8. 2023Time-LLM: Time Series Forecasting by Reprogramming Large Language Modelscited by 680
  9. 2023One Fits All:Power General Time Series Analysis by Pretrained LMcited by 615
  10. 2020Time Series Forecasting With Deep Learning: A Surveycited by 457
  11. 2024TimeMixer: Decomposable Multiscale Mixing for Time Series Forecastingcited by 333
  12. 2018The UEA multivariate time series classification archive, 2018cited by 330
18 more
  1. 2019Unsupervised Scalable Representation Learning for Multivariate Time Seriescited by 267
  2. 2023TSMixer: An All-MLP Architecture for Time Series Forecastingcited by 232
  3. 2021Discrete Graph Structure Learning for Forecasting Multiple Time Seriescited by 169
  4. 2022Less Is More: Fast Multivariate Time Series Forecasting with Light Sampling-oriented MLP Structurescited by 158
  5. 2023LLM4TS: Aligning Pre-Trained LLMs as Data-Efficient Time-Series Forecasterscited by 127
  6. 2023FourierGNN: Rethinking Multivariate Time Series Forecasting from a Pure Graph Perspectivecited by 126
  7. 2024Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Seriescited by 120
  8. 2024SOFTS: Efficient Multivariate Time Series Forecasting with Series-Core Fusioncited by 75
  9. 2021Time Series Prediction using Deep Learning Methods in Healthcarecited by 71
  10. 2025A Comprehensive Survey of Deep Learning for Multivariate Time Series Forecasting: A Channel Strategy Perspectivecited by 58
  11. 2023Multi-resolution Time-Series Transformer for Long-term Forecastingcited by 41
  12. 2023Mitigating Over-smoothing in Transformers via Regularized Nonlocal Functionalscited by 31
  13. 2024MCformer: Multivariate Time Series Forecasting with Mixed-Channels Transformercited by 10
  14. 2024HiMTM: Hierarchical Multi-Scale Masked Time Series Modeling with Self-Distillation for Long-Term Forecastingcited by 8
  15. 2024On using Machine Learning Algorithms for Motorcycle Collision Detectioncited by 4
  16. 2024Optimizing Time Series Forecasting Architectures: A Hierarchical Neural Architecture Search Approachcited by 3
  17. 2024Accelerometer-Based Multivariate Time-Series Dataset for Calf Behavior Classificationcited by 2
  18. 2025Wavy Transformercited by 1

Built on by

Not cited yet.