Research map / Time series / Causal and generative time series
Mean-Field Neural Differential Equations: A Game-Theoretic Approach to Sequence Prediction
Sungwoo Park, Byungseung Kong
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 exponential graphon instead of cosinusoidal graphon (ablation) | MFPs · MIT Humanoid Robot | MSE (x1e-2): 3.91 -> 3.31 | Improved |
| use exponential graphon instead of cosinusoidal graphon (ablation) | MFPs · Beijing Air Quality | MSE (x1e-2): 3.13 -> 2.98 | Improved |
| use exponential graphon instead of cosinusoidal graphon (ablation) | MFPs · MIMIC-II | MSE (x1e-2): 7.51 -> 7.51 | No clear effect |
| replace Mamba with mean-field predictors | sequence predictor · MIT Humanoid Robot | MSE (x1e-2): 5.21 -> 3.31 | Improved |
| replace Mamba with mean-field predictors | sequence predictor · MIMIC-II | MSE (x1e-2): 13.23 -> 7.51 | Improved |
| replace S4 with mean-field predictors for long-term prediction | sequence predictor · EigenWorm (long-term) | MSE (x1e-2): 14.16 -> 12.52 | Improved |
| increase number of sampled predictors N from 1 to 16 (ablation) | MFPs (Exp.) · EigenWorm | MAE | Improved |
| inject Gaussian noise (variance 0.3) into past observations (ablation) | MFPs vs Mamba · EigenWorm | performance degradation | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Contiformer — ContiFormer: Continuous-Time Transformer for Irregular Time Series Modeling (NeurIPS 2023) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Neural Stochastic Differential Games for Time-series Analysis (ICML 2023 · similarity 0.88)
- Sequence Modeling with Spectral Mean Flows (NeurIPS 2025 · similarity 0.86)
- Signed Graph Neural Ordinary Differential Equation for Modeling Continuous-Time Dynamics (AAAI 2024 · similarity 0.85)
- Efficient Training of Neural Stochastic Differential Equations by Matching Finite Dimensional Distributions (ICLR 2025 · similarity 0.85)
- Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions (NeurIPS 2025 · similarity 0.85)
- Universal Time Series Generation with Neural Controlled Differential Equations (NeurIPS 2026 · similarity 0.84)
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.
