Zen and the art of parameter estimation in systems biology

ed appropriately, the specific form of a mathematical model is irrelevant insofar as the numerical optimization of the cost function is concerned, as long as it can evaluate the least-squares deviation of a model from data for a given set of parameters θ . Optimizing an arbitrary nonlinear function of a set of variables is a widespread problem throughout all of science, and accordingly, much algorithmic and development work has been devoted to producing numerical tools capable of carrying out this essential computational task. Numerical optimization is something of an art: there is a vast set of different algorithms that one might possibly make use of, and determining which is most appropriate for a given problem can require a bit of experimentation. Perhaps the most relevant distinguishing feature among different algorithms are those that are capable of identifying global optima and those that make do with finding local optima. In some cases, there can be