OPTIMIZATION 353
B
y A
2
1
0
–1
–2
–2 –1 0 1 2 3
x
Figure 12.9 Function (thin lines) and constraint (thick line)
curves for Example 12.7
Example 12.8. Find the optimal points for f(x,y) = 2 + xy on the circle (x – 2)^2 + (y – 2)^2 = 4.
The Lagrangian becomes
L = 2 + xy + λ[(x – 2)^2 + (y– 2)^2 – 4]
differentiating which yields
∂L/∂x=y + 2λ(x – 2) = 0
∂L/∂y = x + 2λ(y – 2) = 0
solving which we get y = x. Using the constraint, we have
2(y – 2)^2 = 4
ory = 2 ±√2. Thus, the two solution sets are (2 + √2, 2 + √2) and (2 – √2, 2 – √2). The following
summarized solution table is provided along with the function values.
xy λ f(x,y)
(i) 2 + √22 + √ 2 – –^1212 13.66
(ii) 2 – √22 – √ 2 12 –^12 2.34(
The Hessian is computed as
H =
21
12
λ
λ
⎡
⎣
⎢
⎤
⎦
⎥
It can be shown that the Hessian for solution (i) is negative definite and for (ii) it is neither positive
nor negative definite. Thus, solution (i) (point C in Figure 12.10) provides a local maximum while
(ii) (point B) provides a saddle point.
12.3.3 Multivariable Optimization with Inequality Constraints
Consider now, minimizing a function f(X) in n variables with m inequality constraints gi(X)≤ 0,