AI Insight
MPINeuralODE is a framework that improves Neural Ordinary Differential Equations by combining a physics-informed residual term with a Multiple-Initial-Condition (MIC) multiple-shooting curriculum. The physics term constrains the learned vector field while the MIC strategy exposes the model to diverse starting conditions during training, addressing the common failure of Neural ODEs to generalize beyond training trajectories. On the Lotka-Volterra benchmark, the method achieves a 26% reduction in out-of-sample and long-horizon mean squared error compared to baseline Neural ODEs, while maintaining competitive Hamiltonian drift performance.
Why it matters
Improved generalization and long-horizon stability in learned dynamical systems has direct relevance to fields such as ecological modeling, drug pharmacokinetics, climate simulation, and any domain where differential equation models must remain reliable under varied initial conditions. This work offers a structurally principled way to embed physical constraints into data-driven ODE solvers without requiring full mechanistic knowledge of the system.
arXiv:2605.13305v1 Announce Type: cross
Abstract: Neural ordinary differential equations (Neural ODEs) often fit training trajectories while generalizing poorly to unseen initial conditions and long horizons. We propose MPINeuralODE, which combines a soft physics-informed residual with a Multiple-Initial-Condition (MIC) multiple-shooting curriculum whose ingredients are structurally complementary: the physics term anchors the vector-field magnitude on the support that MIC enlarges. We evaluate along three axes: out-of-sample error, long-horizon stability, and Hamiltonian drift, which together expose whether the learned dynamics recover the underlying vector field. On Lotka-Volterra, MPINeuralODE achieves the lowest out-of-sample and long-horizon MSE among data-driven methods, with a 26% reduction over the baseline Neural ODE, while essentially matching the PINN ablation on Hamiltonian drift.