Adaptive optimal input design and parametric estimation of nonlinear dynamical systems: application to neuronal...

Area of Science:

• Computational Neuroscience
• Systems Biology
• Data Assimilation

Background:

• Physical models of biological processes often involve complex nonlinear dynamics.
• Accurate model fitting requires high-quality, dynamically rich experimental data.
• Data assimilation in noisy, non-stationary environments is challenging and data-intensive.

Purpose of the Study:

• To develop an efficient experimental design for faster and more accurate biological model identification.
• To improve parameter estimation in models with limited or noisy measurements.
• To reduce the time and cost associated with biological model fitting.

Main Methods:

• Integration of optimal input design with square root cubature Kalman filters (OID-SCKF).
• Online estimation procedure for adaptive model fitting.
• Benchmarking against standard SCKF methods using nonlinear and neural mass models.

Main Results:

• OID-SCKF demonstrated significantly faster convergence of parameter estimates (up to 1000 ms gain).
• Enhanced model accuracy and reduced dependence on inter-trial noise variability (up to 81% increase).
• Superior performance in identifying parameters for systems with challenging dynamics or limited measurable outputs.

Conclusions:

• OID-SCKF offers a promising approach for rapid and accurate modeling of biological systems, particularly neural dynamics.
• The method is effective even with spatially under-sampled and noisy measurements common in neural engineering.
• Efficient experimental design is crucial for overcoming limitations in biological data assimilation.