Abstract
Computational models have often been used to replicate and explain phenomena and mechanisms in sensorimotor systems. Neuromuscular models are, in fact, computational implementations of hypotheses about the constitutive parts and operant mechanisms of neuromuscular systems. Models for neuromuscular function typically contain multiple elements and their respective parameter values. Often these model elements and their parameter values may be difficult to estimate or measure, describe from first principles, and may vary naturally in the population. Before accepting the result from a simulation, and therefore, the test of a hypothesis, one must explain to the satisfaction of the research community the differences that invariably emerge between model predictions and experimental data and intuition. While these differences can at times be justifiably dismissed as unimportant details, they can also be a product of the validity of the scientific hypothesis being tested, the choice of representation selected for each constitutive element, parameter variability/uncertainty, or even numerical implementation. The use of sensitivity analysis (quantifying the effect of parameter variability on prediction variability) and cross-validation (testing how well a model replicates data not used during its development) are well-established techniques in machine learning and in engineering that are not yet the standard of practice in neuromuscular modeling. I will present some examples of these techniques in the areas of muscle modeling and muscle redundancy that challenge current notions and suggest productive research directions.
I will discuss the implications of this work to systems modeling, neural control, and understanding of neurological pathologies and rehabilitation strategies.
Related publications at:
http://bbdl.usc.edu/Papers/Valero-Cuevas_et_al_RBME_Modeling_2009.pdf