Symbolic function versus linear interpolation

Hi Everyone,

I am solving a bounded ODE using bvp4c, where the objective function is dependent on another fully defined function.

The predefined function I feed into the objective by defining a nest function in the optimization function, for example if I'm solving the ODE

f'(x) = g(x)

where g(x) is my predefined function, then I pass into the solver

function dFdx = odefun(x, F) dFdx = g(x) end

function outg = g(x) %my evaluation of g(x) here end

now, there are two ways I can go about evaluating g(x), I can either evaluate a symbolic function, for example g(x) = e^-2*x, or I can generate a vector of values for g(x) and use outg = interp1(x_vec, g_vec, x).

My question is: does anyone have any intuition on which method - linear interpolation or symbolic evaluation - is computationally superior?

Thanks!

[Darin]

I should specify actually that for the symbolic function method I could really do one of three things:

• use a fit object (cfit object) - specifically a biexponentially - using feval
• use a symbolic function
• pull the parameters of the symbolic function and use an analytic expression.