|
In quantitative fields—such as finance or
physics—practitioners construct mathematical models. Models take many
forms, but they generally comprise a set of assumptions that are
formalized with mathematical equations. Practitioners use models to make
predictions. In finance, a model might be used to predict market
volatility. In physics, it might be used to predict a force of impact.
In some cases, a model will imply a specific formula that
may be used to calculate
such outputs. The practitioner plugs inputs into the formula and
directly obtains the desired output. Such formulas are called
closed-form solutions. A
closed-form solution (or closed form
expression) is any formula that can be evaluated in a finite
number of standard operations.
Consider the formula
 |
[1] |
It is of closed-form because it is expressed as a
finite number of operations: a logarithm, an addition, a division and a
square root. An obvious question is: what is a "standard operation?" There
is no general agreement on this. For example, someone might consider a
Fourier transform to be a standard operation, and someone else might
not. In this regard, the concept of a closed-form solution is informal.
The formula
 |
[2] |
is an infinite sum. Because it entails an infinite
number of standard operations, it is not a closed-form solution. Suppose
we need to evaluate [2]. We can't possibly perform an infinite number of
calculations! One solution is to use our knowledge of calculus to
rearrange [2] as
Formulas [2] and [3] are equivalent. Formula [3] is a
closed-form solution for [2]. In this case, we were fortunate that [2] had
a particularly simple form, which allowed us to find a closed-form
solution for it. Often, we will be confronted with more complicated
formulas, and it will be difficult or impossible to discover a closed-form
solution. In that case, we may seek a
numerical solution instead. What is a numerical solution?
Consider
 |
[4] |
It is a numerical solution for [2]. It approximates [2] by
adding up the first 10,000 terms of that infinite sum.
A numerical solution is any approximation that can be
evaluated in a finite number of standard operations. Closed form
solutions and numerical solutions are similar in that they both can be
evaluated with a finite number of standard operations. They differ in that
a closed-form solution is exact whereas a numerical solution is only
approximate.
Compare numerical solution [4] with closed form solution
[3]. Obviously, [4] will be more difficult to value than [3]. Numerical
solutions often involve extensive numerical calculations. Before the
advent of computers, this could be prohibitive, and practitioners would
strive to find closed-form solutions. Today, computers make it easy to
implement numerical solutions that would have been unimaginable in years
past.
Let's consider a practical example. We want to solve the
polynomial equation:
 |
[5] |
where a, b and c are known constants.
As any high school student knows, there is a closed-form solution to this
problem called the quadratic formula:
 |
[6] |
If she has a good teacher, the student will also know that
there are generalizations of the quadratic formula that solve third order
polynomial equations
 |
[7] |
and fourth order polynomial equations
 |
[8] |
Those solutions are complicated formulas, easily filling
an entire page, but they exist and are exact, so they are closed-form
solutions.
What about fifth order polynomial equations? Is there a
closed-form solution for them? The answer is "no." This is not a matter of
mathematicians looking for but not yet having found a solution.
Mathematicians have proven that no closed-form solutions exist for
fifth or higher order polynomial equations. Higher-order polynomial
equations are solved using numerical methods. A standard solution is
Newton's method. This numerical method starts with an initial "guess" for
a solution. It then uses that value to come up with a second, better
"guess" at a solution. It then performs the same operation on the second
"guess" to come up with a third, even better "guess." Continuing
iteratively in this manner, Newton's method rapidly converges to an
excellent approximation for a solution.
If you are confronted with a math problem and can't find a
closed-form solution, how do you find a numerical solution? Fortunately,
there is a vast literature on this topic that covers a number of standard
approaches for constructing numerical solutions. Those general approaches
are called numerical methods. For
example, Newton's method is a numerical method. When it is applied to a
specific polynomial equation, it becomes a numerical solution for that
particular equation.
The Monte
Carlo method is another standard numerical method. In finance, it is
widely used by risk managers
to solve problems related to quantifying
risk. It is also used by
financial engineers to
price derivatives.
Other numerical methods used by financial engineers are binomial trees and
finite differences.
|