This research proposes a framework for Intermittency Parameter Optimization (IPO) to identify model parameters capable of reproducing specific neurological disease spike patterns, integrating formal characterization of intermittency types, multi-objective optimization, Bayesian optimization, and validation through bifurcation analysis.
Key findings
The IPO framework can reproduce key features of epileptic seizure dynamics, Parkinsonian beta-band oscillations, and bipolar disorder transitions.
Superior pattern fidelity compared to standard gradient-based optimization is achieved.
The method provides interpretable parameter regions corresponding to distinct pathological states.
Limitations & open questions
The framework's scalability with high-dimensional parameter spaces typical of multi-population neural models is a challenge.
Optimized models are rarely validated against bifurcation structure or tested for seizure predictability.