Modeling Conditional Dependencies in Multiagent Trajectories
Research output: Journal contributions › Conference article in journal › Research › peer-review
Standard
In: Proceedings of Machine Learning Research, Vol. 151, 2022, p. 10518-10533.
Research output: Journal contributions › Conference article in journal › Research › peer-review
Harvard
APA
Vancouver
Bibtex
}
RIS
TY - JOUR
T1 - Modeling Conditional Dependencies in Multiagent Trajectories
AU - Rudolph, Yannick
AU - Brefeld, Ulf
N1 - Conference code: 25
PY - 2022
Y1 - 2022
N2 - We study modeling joint densities over sets of random variables (next-step movements of multiple agents) which are conditioned on aligned observations (past trajectories). For this setting, we propose an autoregressive approach to model intra-timestep dependencies, where distributions over joint movements are represented by autoregressive factorizations. In our approach, factors are randomly ordered and estimated with a graph neural network to account for permutation equivariance, while a recurrent neural network encodes past trajectories. We further propose a conditional two-stream attention mechanism, to allow for efficient training of random factorizations. We experiment on trajectory data from professional soccer matches and find that we model low frequency trajectories better than variational approaches.
AB - We study modeling joint densities over sets of random variables (next-step movements of multiple agents) which are conditioned on aligned observations (past trajectories). For this setting, we propose an autoregressive approach to model intra-timestep dependencies, where distributions over joint movements are represented by autoregressive factorizations. In our approach, factors are randomly ordered and estimated with a graph neural network to account for permutation equivariance, while a recurrent neural network encodes past trajectories. We further propose a conditional two-stream attention mechanism, to allow for efficient training of random factorizations. We experiment on trajectory data from professional soccer matches and find that we model low frequency trajectories better than variational approaches.
KW - Informatics
KW - Business informatics
UR - http://www.scopus.com/inward/record.url?scp=85163100847&partnerID=8YFLogxK
M3 - Conference article in journal
VL - 151
SP - 10518
EP - 10533
JO - Proceedings of Machine Learning Research
JF - Proceedings of Machine Learning Research
SN - 2640-3498
T2 - The 25th International Conference on Artificial Intelligence and Statistics - AISTATS 2022
Y2 - 28 March 2022 through 30 March 2022
ER -