# Interesting optimization problem

• May 28th 2010, 12:50 PM
luispipe
Interesting optimization problem
Hi,

I would like to know if the following optimization problem has any application (or has been used to solve anything):

$
\begin{array}{*{20}{c}}
{{\text{maximize}}} & {J=\sum\limits_{j = 1}^n {{a_j}ln {p_j} } } & {} \\
{{\text{subject to}}} & {\sum\limits_{j = 1}^n {{p_j} = 1} } & {} \\
{} & {{p_i} \geqslant 0,} & {i = 1, \ldots ,n} \\
\end{array}
$

where $a_i$ is a positive constant for i=1,...,n

• May 31st 2010, 07:08 AM
GeoC
Looks remarkably similar to Boltzman's characterization for the distribution of the elements of a system across states $j$, where $P_j$is the probability of occupancy of state $j$. This is subject to the constraint that $\sum P_j = 1$. In the case of Boltzmann, $a_j = P_j$
• May 31st 2010, 09:28 AM
GeoC
I took a stab at this problem, and find the optimum to arise when $a_j =
\lambda P_j$
and $\sum a_j = \lambda$. Since $\lambda$ is an arbitrary scalar, simply take $\lambda = 1$ and you arrive at Boltzmann's equation for H.

Thus $J_{max} = \sum_j P_jlnP_j$
with $\sum_j P_j = 1$
• May 31st 2010, 12:09 PM
SpringFan25
Certain economics problems could be expressed in this way.

If a person must spread a fixed resource (say, Time), over J activities, and the Utility derived from each activity is equal to ${a_k}ln(P_k)$ Where $P_k$ is the proportion of your time that you spend doing activity k. You would maximise

The constraints would come from:
All time must be allocated
All time must be positive

It would be stupid to assume anyone had this utility function of course, but that never stopped economists before.