ABSTRACT
This paper analyzes the stability of Neural Ordinary Differential Equations (Neural ODEs) trajectories under noisy EEG initialization, deriving stability bounds and proposing novel regularization strategies for EEG applications.
PAPER · PDF
Loading PDF...
Key findings
Derives Lipschitz-based stability bounds for EEG applications.
Establishes Lyapunov stability criteria for Neural ODEs under noisy initialization.
Proposes novel initialization strategies and noise-aware regularization techniques.
Develops practical solver selection guidelines based on stability region analysis.
Limitations & open questions
The study focuses on theoretical analysis and requires further empirical validation on diverse EEG datasets.