Interactive sequential generative models for team sports

Research output: Journal contributionsJournal articlesResearchpeer-review

Standard

Interactive sequential generative models for team sports. / Fassmeyer, Dennis; Cordes, Moritz; Brefeld, Ulf.
In: Machine Learning, Vol. 114, No. 2, 38, 02.2025.

Research output: Journal contributionsJournal articlesResearchpeer-review

Harvard

APA

Vancouver

Fassmeyer D, Cordes M, Brefeld U. Interactive sequential generative models for team sports. Machine Learning. 2025 Feb;114(2):38. doi: 10.1007/s10994-024-06648-2

Bibtex

@article{4cc4188e57574979abe1c5f5796a3e6f,
title = "Interactive sequential generative models for team sports",
abstract = "Understanding spatiotemporal coordination of players in team sports is key to movement models, pattern detection, and computational tactics. Existing generative models propose to capture all stochasticity by a single latent variable and may suffer from entangled representations, or aim to uncover interaction structures of players but then do not focus on their generative ability. As a remedy, we propose a hierarchical latent variable model for predicting trajectories of multiple players. In the generative model, both, discrete role assignments and a latent interaction graph are sampled to allow for different models in subsequent node updates and message passing operations between nodes, where standard Gaussian latent variables are employed per agent and timestep. We cast our approach as a variational autoencoder that provides a disentangled latent space to capture variability in team sport movements and propose a neural architecture for its optimization. We empirically evaluate our approach on tracking data from basketball and soccer and observe that our contribution outperforms the state-of-art in all experiments.",
keywords = "Generative models, Graph recurrent neural networks, Soccer, Socio-temporal dependencies, Trajectory forecasting, Informatics",
author = "Dennis Fassmeyer and Moritz Cordes and Ulf Brefeld",
note = "Publisher Copyright: {\textcopyright} The Author(s) 2025.",
year = "2025",
month = feb,
doi = "10.1007/s10994-024-06648-2",
language = "English",
volume = "114",
journal = "Machine Learning",
issn = "0885-6125",
publisher = "Springer Netherlands",
number = "2",

}

RIS

TY - JOUR

T1 - Interactive sequential generative models for team sports

AU - Fassmeyer, Dennis

AU - Cordes, Moritz

AU - Brefeld, Ulf

N1 - Publisher Copyright: © The Author(s) 2025.

PY - 2025/2

Y1 - 2025/2

N2 - Understanding spatiotemporal coordination of players in team sports is key to movement models, pattern detection, and computational tactics. Existing generative models propose to capture all stochasticity by a single latent variable and may suffer from entangled representations, or aim to uncover interaction structures of players but then do not focus on their generative ability. As a remedy, we propose a hierarchical latent variable model for predicting trajectories of multiple players. In the generative model, both, discrete role assignments and a latent interaction graph are sampled to allow for different models in subsequent node updates and message passing operations between nodes, where standard Gaussian latent variables are employed per agent and timestep. We cast our approach as a variational autoencoder that provides a disentangled latent space to capture variability in team sport movements and propose a neural architecture for its optimization. We empirically evaluate our approach on tracking data from basketball and soccer and observe that our contribution outperforms the state-of-art in all experiments.

AB - Understanding spatiotemporal coordination of players in team sports is key to movement models, pattern detection, and computational tactics. Existing generative models propose to capture all stochasticity by a single latent variable and may suffer from entangled representations, or aim to uncover interaction structures of players but then do not focus on their generative ability. As a remedy, we propose a hierarchical latent variable model for predicting trajectories of multiple players. In the generative model, both, discrete role assignments and a latent interaction graph are sampled to allow for different models in subsequent node updates and message passing operations between nodes, where standard Gaussian latent variables are employed per agent and timestep. We cast our approach as a variational autoencoder that provides a disentangled latent space to capture variability in team sport movements and propose a neural architecture for its optimization. We empirically evaluate our approach on tracking data from basketball and soccer and observe that our contribution outperforms the state-of-art in all experiments.

KW - Generative models

KW - Graph recurrent neural networks

KW - Soccer

KW - Socio-temporal dependencies

KW - Trajectory forecasting

KW - Informatics

UR - http://www.scopus.com/inward/record.url?scp=85217779916&partnerID=8YFLogxK

U2 - 10.1007/s10994-024-06648-2

DO - 10.1007/s10994-024-06648-2

M3 - Journal articles

VL - 114

JO - Machine Learning

JF - Machine Learning

SN - 0885-6125

IS - 2

M1 - 38

ER -

Recently viewed

Researchers

  1. Zhiyong Xie

Publications

  1. SoilTemp: A global database of near-surface temperature
  2. Formative assessment in inclusive mathematics education in secondary schools
  3. Comparison of EKF and TSO for Health Monitoring of a Textile-Based Heater Structure and its Control
  4. Avoiding irreversible change
  5. Optimization of waste management systems by integrating material fluxes, agents and regulatory mechanisms
  6. Second-order SMC with disturbance compensation for robust tracking control in PMSM applications
  7. Foreign bias in institutional portfolio allocation
  8. How can problems be turned into something good? The role of entrepreneurial learning and error mastery orientation
  9. Transformation products in the water cycle and the unsolved problem of their proactive assessment
  10. Tree species identity, canopy structure and prey availability differentially affect canopy spider diversity and trophic composition
  11. The case of the composite Higgs
  12. A latent state-trait analysis of current achievement motivation across different tasks of cognitive ability
  13. An empirically grounded ontology for analyzing IT-based interventions in business ecosystems
  14. Learner characteristics and information processing in multimedia learning
  15. Correction to: Operative communication: project Cybersyn and the intersection of information design, interface design, and interaction design (AI & SOCIETY, (2022), 10.1007/s00146-021-01346-2)
  16. Visual-Inertial Navigation Systems and Technologies
  17. Controlling a Bank Model Economy by Sliding Mode Control with Help of Kalman Filter
  18. Lyapunov stability analysis to set up a saturating PI controller with anti-windup for a mass flow system
  19. Editorial: Governance for Sustainable Development in the Face of Ambivalence, Uncertainty and Distributed Power: an Introduction
  20. Improving mechanical properties of chip-based aluminum extrudates by integrated extrusion and equal channel angular pressing (iECAP)
  21. Adding the “e-” to Learning for Sustainable Development
  22. Integration of laboratory experiments into introductory electrical engineering courses
  23. Organizational practices for the aging workforce
  24. Conceptualizing community in energy systems
  25. Effectiveness of an online recovery training for employees exposed to blurred boundaries between work and non-work
  26. Understanding and managing post-acquisition integration as change process