The representation of the process,
, is the sequence of random variables
. The choice basis of
is unrestricted. Of particular interest is to
restrict the basis functions to those which make the
uncorrelated random variables.
When this requirement is satisfied, the resulting representationof
is termed the
Karhunen-Love expansion. Mathematically, we require
. This requirement can be expressed in terms of the
correlation function of
.
As
is given by
our requirement becomes
Simple manipulations result in the expression
When
, the quantity
is just the variance of
. Our requirement is obtained by satisfying
or
where
Furthermore, this requirement must hold for each
which differs from the choice of
. A choice of a function
satisfying this requirement is a function which is
proportional to
:
. Therefore,
The
which allow the representation of
to be a sequence of uncorrelated random variables must
satisfy this integral equation. This type of equation occursoften in applied mathematics; it is termed the
eigenequation . The sequences
and
are the eigenfunctions and eigenvalues of
, the covariance function of
. It is easily verified that:
This result is termed
Mercer's Theorem .
The approach to solving for the eigenfunction and eigenvalues of
is to convert the integral equation into an ordinary
differential equation which can be solved. This approach isbest illustrated by an example.
. The eigenequation can be written in this case as
Evaluating the first derivative of this expression,
or
Evaluating the derivative of the last expression yields the
simple equation
This equation has a general solution of the form
. It is easily seen that
must be zero. The amplitude
is found by requiring
. To find
, one
must return to the original integral equation. Substituting, wehave
After some manipulation, we find that
or
Therefore,
and we have
The Karhunen-Love expansion has several important properties.
- The eigenfunctions of a positive-definite covariance
function constitute a complete set. One can easily show thatthese eigenfunctions are also mutually orthogonal with respect
to both the usual inner product and with respect to the innerproduct derived from the covariance function.
- If
Gaussian,
are Gaussian
random variables. As the random variables
are uncorrelated and Gaussian, the
comprise a sequence of statistically independent
random variables.
- Assume
: the stochastic process
is white. Then
for all
. Consequently, if
, this constraint equation is satisfied
no matter what choice is made for the orthonormal set
. Therefore, the representation of white, Gaussian
processes consists of a sequence of statisticallyindependent, identically-distributed (mean zero and variance
) Gaussian random variables. This example
constitutes the simplest case of the Karhunen-Love
expansion.