Kernel Ordinary Differential Equations.

Area of Science:

• Computational Biology
• Applied Mathematics
• Statistical Modeling

Background:

• Ordinary differential equations (ODEs) are fundamental for modeling dynamic systems in science.
• Existing methods often require known functional forms or linearity, limiting their applicability.
• Noisy observations pose significant challenges for accurate ODE estimation.

Purpose of the Study:

• To develop a flexible, kernel-based approach for estimating and inferring ODEs from noisy data.
• To address limitations of existing methods by not assuming known functional forms or linearity.
• To enable the selection of relevant functional components and provide confidence intervals for estimated trajectories.

Main Methods:

• A reproducing kernel-based method is proposed for ODE estimation.
• Sparse estimation techniques are employed for functional selection.
• The approach builds upon the smoothing spline analysis of variance (SS-ANOVA) framework.
• Confidence intervals are constructed for estimated signal trajectories.

Main Results:

• The method achieves estimation optimality and selection consistency in both low- and high-dimensional settings.
• It effectively handles unknown, non-linear, and interacting functional forms in ODEs.
• The approach extends the capabilities of existing SS-ANOVA methods.

Conclusions:

• The proposed kernel ODE method offers a powerful and flexible tool for modeling complex dynamical systems.
• It provides robust estimation and inference for ODEs even with noisy and limited data.
• This work advances the state-of-the-art in statistical inference for differential equations.