Research map / Probabilistic and scientific modeling / Dynamical systems
Unsupervised Learning of Lagrangian Dynamics from Images for Prediction and Control
Yaofeng Desmond Zhong, Naomi Leonard
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace coordinate-aware VAE with traditional VAE (ablation) | Lagrangian+caVAE · CartPole test | average pixel MSE (x1e-3): 2.81 -> 10.30 | Hurt |
| replace coordinate-aware VAE with traditional VAE (ablation) | Lagrangian+caVAE · Pendulum test | average pixel MSE (x1e-3): 1.83 -> 1.56 | No clear effect |
| replace coordinate-aware VAE with traditional VAE (ablation) | Lagrangian+caVAE · Acrobot test | average pixel MSE (x1e-3): 3.14 -> 10.78 | Hurt |
| replace Lagrangian latent dynamics with MLP dynamics (ablation) | Lagrangian+caVAE · CartPole test | average pixel MSE (x1e-3): 2.81 -> 14.97 | Hurt |
| replace Lagrangian latent dynamics with MLP dynamics (ablation) | Lagrangian+caVAE · Acrobot test | average pixel MSE (x1e-3): 3.14 -> 12.14 | Hurt |
| replace Lagrangian latent dynamics with MLP dynamics (ablation) | Lagrangian+caVAE · Pendulum test | average pixel MSE (x1e-3): 1.83 -> 1.92 | No clear effect |
| replace coordinate-aware decoder with black-box MLP (ablation) | Lagrangian+caVAE · interpretability of coordinates | qualitative | Hurt |
| replace coordinate-aware encoder with black-box MLP (ablation) | Lagrangian+caVAE · interpretability of coordinates | qualitative | Hurt |
| replace coordinate-aware VAE with coordinate-aware AE (ablation) | Lagrangian+caVAE · CartPole | qualitative | Hurt |
| replace the model with PAIG (AE with velocity estimator) (ablation) | Lagrangian+caVAE · long term prediction | qualitative | Hurt |
Closest papers by meaning
- A Riemannian Framework for Learning Reduced-order Lagrangian Dynamics (ICLR 2025 · similarity 0.89)
- Simplifying Hamiltonian and Lagrangian Neural Networks via Explicit Constraints (NeurIPS 2020 · similarity 0.89)
- Learned Lagrangian Models of PDEs via Euler–Lagrange Residual Minimization (NeurIPS 2026 · similarity 0.87)
- Learning Physics From Video: Unsupervised Physical Parameter Estimation for Continuous Dynamical Systems (CVPR 2025 · similarity 0.87)
- Discovering Phase Space Structure in Learned Hamiltonian Systems (NeurIPS 2026 · similarity 0.85)
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
