Graph Conditional Variational Models: Too Complex for Multiagent Trajectories?

Publikation: Beiträge in ZeitschriftenKonferenzaufsätze in FachzeitschriftenForschungbegutachtet

Authors

Recent advances in modeling multiagent trajectories combine graph architectures such as graph neural networks (GNNs) with conditional variational models (CVMs) such as variational RNNs (VRNNs). Originally, CVMs have been proposed to facilitate learning with multi-modal and structured data and thus seem to perfectly match the requirements of multi-modal multiagent trajectories with their structured output spaces. Empirical results of VRNNs on trajectory data support this assumption. In this paper, we revisit experiments and proposed architectures with additional rigour, ablation runs and baselines. In contrast to common belief, we show that prior results with CVMs on trajectory data might be misleading. Given a neural network with a graph architecture and/or structured output function, variational autoencoding does not seem to contribute statistically significantly to empirical performance. Instead, we show that well-known emission functions do contribute, while coming with less complexity, engineering and computation time.
OriginalspracheEnglisch
ZeitschriftProceedings of Machine Learning Research
Jahrgang137
Seiten (von - bis)136-147
Anzahl der Seiten12
ISSN2640-3498
PublikationsstatusErschienen - 2020
VeranstaltungVirtual NeurIPS 2020: Neural Information Processing Systems Online Conference 2020 - digital
Dauer: 06.12.202012.12.2020
Konferenznummer: 34
https://neurips.cc/virtual/2020/public/index.html
https://proceedings.mlr.press/v137/

Bibliographische Notiz

Publisher Copyright:
© Proceedings of Machine Learning Research 2020.

Zugehörige Aktivitäten

Links

Zuletzt angesehen

Publikationen

  1. Analysis of Complexity Reduction in Kalman Filters Through Decoupling Control With Chattered Inputs in PMSM
  2. A Quadrant Approach of Camera Calibration Method for Depth Estimation Using a Stereo Vision System
  3. Inversion of Fuzzy Neural Networks for the Reduction of Noise in the Control Loop for Automotive Applications
  4. On the Power and Performance of a Doubly Latent Residual Approach to Explain Latent Specific Factors in Multilevel-Bifactor-(S-1) Models
  5. Noise level estimation using haar wavelet packet trees for sensor robust outlier detection
  6. Closed-loop control of product geometry by using an artificial neural network in incremental sheet forming with active medium
  7. Comments on “Modular-Controller-Design-Based Fast Terminal Sliding Mode for Articulated Exoskeleton Systems”
  8. Multi-Parallel Sending Coils for Movable Receivers in Inductive Charging Systems
  9. Learning Rotation Sensitive Neural Network for Deformed Objects' Detection in Fisheye Images
  10. Design and Control of an Inductive Power Transmission System with AC-AC Converter for a Constant Output Current
  11. Vision-Based Deep Learning Algorithm for Detecting Potholes
  12. Lyapunov stability analysis to set up a PI controller for a mass flow system in case of a non-saturating input
  13. Ant colony optimization algorithm and artificial immune system applied to a robot route
  14. A geometric algorithm for the output functional controllability in general manipulation systems and mechanisms
  15. A geometric approach for controlling an electromagnetic actuator with the help of a linear Model Predictive Control
  16. Analyzing math teacher students' sensitivity for aspects of the complexity of problem oriented mathematics instruction
  17. For a return to the forgotten formula: 'Data 1 + Data 2 > Data 1'
  18. Modeling Conditional Dependencies in Multiagent Trajectories
  19. Data-Driven flood detection using neural networks
  20. An evaluation of BPR methodologies adopting NIMSAD: A systematic framework for understanding and evaluating methodologies
  21. Problem solving in mathematics education
  22. Multi-view learning with dependent views
  23. Don’t underestimate the problems of user centredness in software development projectsthere are many!?
  24. Control of an Electromagnetic Linear Actuator Using Flatness Property and Systems Inversion
  25. A PHENOMENOGRAPHICAL STUDY OF CHILDRENS’ SPATIAL THOUGHT WHILE USING MAPS IN REAL SPACES
  26. Parking space management through deep learning – an approach for automated, low-cost and scalable real-time detection of parking space occupancy
  27. Automated Invoice Processing: Machine Learning-Based Information Extraction for Long Tail Suppliers
  28. On the Functional Controllability Using a Geometric Approach together with a Decoupled MPC for Motion Control in Robotino
  29. Nonlinear PD fault-tolerant control for dynamic positioning of ships with actuator constraints