Research map / Time series / Causal and generative time series
Homotopy-based training of NeuralODEs for accurate dynamics discovery
Joon-Hyuk Ko, Hankyul Koh, Nojun Park, Wonho Jhe
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 |
|---|---|---|---|
| use homotopy training instead of vanilla gradient descent | NeuralODE · Lotka-Volterra Black Box | interpolation MSE (x1e-2): 1.76 -> 0.78 | Improved |
| use homotopy training instead of vanilla gradient descent | NeuralODE · Lotka-Volterra Gray Box | interpolation MSE (x1e-2): 27.3 -> 0.72 | Improved |
| use homotopy training instead of vanilla gradient descent | NeuralODE · Double Pendulum Second Order | interpolation MSE (x1e-2): 1.03 -> 0.18 | Improved |
| use homotopy training instead of vanilla gradient descent | NeuralODE · Lorenz Black Box | interpolation MSE (x1e-2): 259 -> 18.6 | Improved |
| use multiple shooting instead of homotopy | NeuralODE · Lorenz Black Box | interpolation MSE (x1e-2): 18.6 -> 12.9 | Mixed |
| use multiple shooting instead of vanilla for extrapolation | NeuralODE · Double Pendulum | extrapolation MSE | Hurt |
| use homotopy training instead of vanilla | NeuralODE · all benchmarks | epochs to minimum interpolation MSE: 90%-40% reduction | Improved |
| use homotopy training instead of vanilla | NeuralODE · Lotka-Volterra Black Box | best train epochs: (3932, 3818, 3134) -> (299, 208, 227) | Improved |
| increase training data length (ablation) | vanilla gradient descent · Lotka-Volterra | interpolation/extrapolation MSE | Hurt |
| increase data length (ablation) | homotopy and multiple shooting · Lotka-Volterra | MSE | No clear effect |
| decrease model size with homotopy or multiple shooting (ablation) | NeuralODE · Lotka-Volterra | MSE | No clear effect |
| increase data sparsity and noise (ablation) | homotopy training · Lotka-Volterra | MSE | No clear effect |
Closest papers by meaning
- ControlSynth Neural ODEs: Modeling Dynamical Systems with Guaranteed Convergence (NeurIPS 2024 · similarity 0.89)
- Semi-supervised Learning of Dynamical Systems with Neural Ordinary Differential Equations: A Teacher-Student Model Approach (AAAI 2024 · similarity 0.89)
- Learn Singularly Perturbed Solutions via Homotopy Dynamics (ICML 2025 · similarity 0.88)
- DISCO: learning to DISCover an evolution Operator for multi-physics-agnostic prediction (ICML 2025 · similarity 0.88)
- How Deep Do We Need: Accelerating Training and Inference of Neural ODEs via Control Perspective (ICML 2024 · similarity 0.87)
- From Fourier to Neural ODEs: Flow Matching for Modeling Complex Systems (ICML 2024 · similarity 0.87)
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.
