Semi-Supervised Generative Models for Multi-Agent Trajectories

Publikation: Beiträge in SammelwerkenAufsätze in KonferenzbändenForschungbegutachtet

Authors

Analyzing the spatiotemporal behavior of multiple agents is of great interest to many communities. Existing probabilistic models in this realm are formalized either in an unsupervised framework, where the latent space is described by discrete or continuous variables, or in a supervised framework, where weakly preserved labels add explicit information to continuous latent representations. To overcome inherent limitations, we propose a novel objective function for processing multi-agent trajectories based on semi-supervised variational autoencoders, where equivariance and interaction of agents are captured via customized graph networks. The resulting architecture disentangles discrete and continuous latent effects and provides a natural solution for injecting expensive domain knowledge into interactive sequential systems. Empirically, our model not only outperforms various state-of-the-art baselines in trajectory forecasting, but also learns to effectively leverage unsupervised multi-agent sequences for classification tasks on interactive real-world sports datasets.

OriginalspracheEnglisch
TitelAdvances in Neural Information Processing Systems 35 : 36th Conference on Neural Information Processing Systems (NeurIPS 2022)
HerausgeberS. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, A. Oh
Anzahl der Seiten15
Band48
ErscheinungsortRed Hook
VerlagCurran Associates
Erscheinungsdatum2022
Seiten37267-37281
ISBN (Print)978-1-7138-7108-8
ISBN (elektronisch)978-1-7138-7312-9
PublikationsstatusErschienen - 2022
Veranstaltung36th Conference on Neural Information Processing Systems - NeurIPS 2022 - New Orleans Convention Center, New Orleans, USA / Vereinigte Staaten
Dauer: 28.11.202209.12.2022
Konferenznummer: 36
https://nips.cc/Conferences/2022

Bibliographische Notiz

Funding Information:
We would like to thank the German Football Association (DFB) for providing data for this study.

Publisher Copyright:
© 2022 Neural information processing systems foundation. All rights reserved.

Zuletzt angesehen

Publikationen

  1. Instruments for research on transition. Applied methods and approaches for exploring the transition of young care leavers to adulthood
  2. Leveraging the macro-level environment to balance work and life
  3. Robustness of coherent sets computations
  4. Structural ambidexterity, transition processes, and integration trade‐offs: a longitudinal study of failed exploration
  5. Putting adaptive planning into practice: A meta-analysis of current applications
  6. How secondary-school students deal with issues of sustainable development in class*
  7. Towards greener and sustainable ionic liquids using naturally occurring and nature-inspired pyridinium structures
  8. Competition in fragmented markets
  9. Tree diversity and nectar composition affect arthropod visitors on extrafloral nectaries in a diversity experiment
  10. The impact of partially missing communities on the reliability of centrality measures
  11. Flexibility of industrial material flow networks
  12. Editorial: Governance for Sustainable Development in the Face of Ambivalence, Uncertainty and Distributed Power: an Introduction
  13. Performance of the Chemcatcher ® passive sampler when used to monitor 10 polar and semi-polar pesticides in 16 Central European streams, and comparison with two other sampling methods
  14. Beyond Structural Adjustment
  15. Introduction to General Ecology
  16. The effects of an Internet based self-help course for reducing panic symptoms-Don't Panic Online
  17. Analytical and Experimental Performance Analysis of Enhanced Wake-Up Receivers Based on Low-Power Base-Band Amplifiers
  18. Calibrated Passive Sampling - Multi-plot Field Measurements of NH3 Emissions with a Combination of Dynamic Tube Method and Passive Samplers