next up previous
Next: Global Extrema Up: Applications of Derivatives Previous: Example of Constrained Optimization

Interpretation of the Lagrange Multiplier

To interpret the Lagrange multiplier $\lambda$ consider the case where we have found that (x0,y0) is the optimum value of f(x,y) subject to the constraint $g(x,y)
\le c$. Let's examine what happens if we vary the value of c (increase/decrease our budget.)

\begin{displaymath}
\frac{df}{fc} = \frac{\partial f}{\partial x} \frac{dx_0}{dc} +
\frac{\partial f}{\partial y} \frac{dy_0}{dc}\end{displaymath} (14)

Using the fact that $\mbox{grad}f = \lambda \mbox{grad}g$ at the point (x0,y0) we can rewrite this as

\begin{displaymath}
\frac{df}{dc} = \lambda \left( \frac{\partial g}{\partial x}...
 ...g}{\partial y} \frac{dy_0}{dc} \right) =
\lambda \frac{dg}{dc}.\end{displaymath} (15)

However at the critical point, g(x0(c), y0(c)) = c so dg/dc = 1 and hence, $df/dc = \lambda$. Thus $\lambda$ represents ``how much more bang you get for your buck.'' A more mathematical statement of this is

\begin{displaymath}
\frac{d}{dc} f(x_0(c), y_0(c)) = \lambda.\end{displaymath} (16)


Vector Calculus
12/6/1997