# Numerical solution

Last time, I asked this question

but after discussing with some friends, I have given up finding closed-form solutions. Now I have a simpler question.Let $g_i: i=1,2$ be $C^2 =C^{2}(-\infty,\infty)$ functions such that $0\leq g_1(x)\leq C_1$ and $0\leq g_2(x)\leq e^{x}C_2$ for some FIXED constants $C_1,C_2$. If $f_i : i=1,2$ are $C^2$ functions, let $\lambda$ be a fixed positive constants then define

$Af_i(x)=(\mu -\frac{\sigma^2}{2})\frac{\partial f_i(x)}{\partial x}+\frac{\sigma^2}{2}\frac{\partial^2 f_i(x)}{\partial x^2} +\lambda(g_i(x)-f_i(x))$.

I am interested in finding NUMERICAL SOLUTIONS( i.e finding $f_1, f_2$ numerically) of the following systems of equations

$\min \{ \rho f_1(x) - Af_1(x), f_1(x)- f_2(x) +e^x D_1\} =0$

$\min \{ \rho f_2(x) - Af_2(x), f_2(x)- f_1(x) - e^x D_2\} =0$

where $\mu,\sigma,\rho, D_1>D_2$ are fixed positive constants

It would be highly appreciated if some one could give me some ideas on finding a scheme to solve this problem or at least a reference to this type of problem

-
Please do not use the word analytic when you mean closed-form. I do not know what kind of training would cause a person to use the word analytic in this context, but there is a precise meaning in mathematics in the English language for the word. –  Will Jagy Jul 27 '11 at 22:15
thanks Will, I will be careful next time –  Steven Jul 27 '11 at 22:23
@Will: people in computer science sometimes overload 'analytic' to mean 'closed-form'; a further semi-acceptable stretch is to use it to mean 'exactly computable using a (finite) algorithm'...but yes, this overloading is imprecise and should be avoided (I myself am guilty of this abuse) –  Suvrit Jul 28 '11 at 0:10