Upload
vulien
View
213
Download
0
Embed Size (px)
Citation preview
Unconstrained Optimization(Maximization)
• Ingredients
x choice variable (control v., instrument, decision v.)
X opportunity set: x ∈ X
unconstrained problems: X = Rconstrained problems: X ⊂ R
y objective (related to x)
objective function: y = f (x)
• choose x such as to max f (x)
& observe x ∈ X
• solution: x∗ ∈ X f (x∗) ≥ f (x) for all x ∈ X
x∗ “optimizer” or “maximizer”
• f ′(x) = 0 necessary condition for interior optimum
c© Ronald Wendner Summary M4M-II-1 v1.3
• points to note
– necessary vs. sufficient
– inner vs. boundary solution
– existence
• Weierstrass existence theorem
f (x) continuous
X nonempty, closed, bounded
– sufficient!
• local vs. global solutions
– global solution: x∗ ∈ X f (x∗) ≥ f (x) for all
x ∈ X
– local solution: x∗ ∈ X f (x∗) ≥ f (x) for all
x ∈ Nε(x)
– every global solution = local solution.
– sufficient condition(s) for x∗ to be global sol.?
c© Ronald Wendner Summary M4M-II-2 v1.3
• Local-global theorem
– f (x) concave function & X is convex set then:
(i) every local opt. is global opt.
(ii) set of global opt. is convex
– strict quasiconcavity is sufficient as well
– uniqueness
f (x) strictly (quasi)concave
• Second order conditions (suppose f ′(x) = 0)
– x is a solution ⇒ f ′′(x) ≤ 0 (necessary SOC)
– f ′(x) = 0 and f ′′(x) < 0 ⇒ x is a solution
(sufficient SOC)
c© Ronald Wendner Summary M4M-II-3 v1.3
Sufficient condition ⇒ Necessary condition
1. x∗ is maximizer ⇒ f ′(x∗) = 0, f ′′(x∗) ≤ 0
2. f ′(x∗) = 0, f ′′(x∗) < 0 ⇒ x∗ is maximizer
c© Ronald Wendner Summary M4M-II-4 v1.3
Several Choice Variables
• Ingredients
N choice variables: x ≡ (x1, x2, ..., xN)′
X opportunity set: x ∈ X
unconstrained problems: X = RN
constrained problems: X ⊂ RN
objective (related to x)
objective function: f (x)
• choose x ∈ X such as to max f (x)
• 2 complications
– necessary FOC: @ f ′(x)
– nec., suff. SOC: @ f ′′(x)
c© Ronald Wendner Summary M4M-II-5 v1.3
• FOC: derivative → partial derivative
– fi(x) ≡ ∂f(x)∂xi
= 0 for i = 1, ..., N
– N necessary FOC
• SOC: second order partial derivatives
– necessary: f (x) concave ⇔ H negative semidef-
inite
– sufficient: f (x) strictly concave ⇔ H negative
definite
→ linear algebra sessions
c© Ronald Wendner Summary M4M-II-6 v1.3