Introduction to Unconstrained Optimization - Scilab
Introduction to Unconstrained Optimization - Scilab
Introduction to Unconstrained Optimization - Scilab
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
for all x, y ∈ C, then f is convex over the convex set C. This proof is given in a slightly more<br />
complex form in [2], chapter 10 ”Conditions d’optimalité”, section 10.1.2 ”Différentiabilité”.<br />
Assume that x, y are two points in the convex set C. Since C is a convex set, the point<br />
x + θ(y − x) = θy + (1 − θ)x is also in C, for all θ so that 0 ≤ θ ≤ 1. Let us define the following<br />
function<br />
φ(θ) = f(x + θ(y − x)), (84)<br />
for all θ so that 0 ≤ θ ≤ 1.<br />
The idea of the proof is based on the fact that the convexity of φ with respect <strong>to</strong> θ ∈ [0, 1] is<br />
equivalent <strong>to</strong> the convexity of f with respect <strong>to</strong> x ∈ C. We will prove that<br />
φ ′ (t) − φ ′ (s) ≥ 0, (85)<br />
for all real values s and t so that t ≥ s. That inequality means that φ ′ is an increasing function,<br />
which property is associated with a convex function. Then we will integrate the inequality over<br />
well chosen ranges of s and t so that the convexity of φ appear.<br />
To prove the inequality 85, let us define two real values s, t ∈ R. We apply the inequality 83<br />
with the two points x + t(y − x) and x + s(y − x). By the convexity of C, these points are in C<br />
so that we get<br />
[g(x + t(y − x)) − g(x + s(y − x))] T (x + t(y − x) − x − s(y − x) ≥ 0. (86)<br />
The left hand side can be simplified in<strong>to</strong> (t−s)(φ ′ (t)−φ ′ (s)) ≥ 0. For t ≥ s, the previous inequality<br />
leads <strong>to</strong> 85.<br />
We now integrate the inequality 85 for t ∈ [θ, 1] and s ∈ [0, θ] so that the order t ≥ s is satisfied<br />
and we get<br />
∫ ∫<br />
(φ ′ (t) − φ ′ (s)) dtds ≥ 0. (87)<br />
t∈[θ,1]<br />
s∈[0,θ]<br />
We first compute the following two integrals<br />
∫ ∫<br />
φ ′ (t)dtds =<br />
s∈[0,θ]<br />
t∈[θ,1]<br />
∫<br />
s∈[0,θ]<br />
(φ(1) − φ(θ))ds (88)<br />
= θ(φ(1) − φ(θ)), (89)<br />
and<br />
∫ ∫<br />
∫<br />
φ ′ (s)dsdt = (φ(θ) − φ(0))dt (90)<br />
t∈[θ,1] s∈[0,θ]<br />
t∈[θ,1]<br />
= (1 − θ)(φ(θ) − φ(0)). (91)<br />
By plugging the two integrals 89 and 91 in<strong>to</strong> the inequality 87, we get<br />
which simplifies <strong>to</strong><br />
θ(φ(1) − φ(θ)) − (1 − θ)(φ(θ) − φ(0)) ≥ 0, (92)<br />
θφ(1) + (1 − θ)φ(0) − φ(θ) ≥ 0. (93)<br />
which proves that φ is convex. If we plug the definition 84 of φ and the equalities φ(0) = f(x) and<br />
φ(1) = f(y), we find the convexity inequality for f<br />
which concludes the proof.<br />
θf(y) + (1 − θ)f(x) ≤ f(θy + (1 − θ)x), (94)<br />
Answer of Exercise 2.6 (Hessian of Rosenbrock’s function) Let us prove that the Hessian<br />
matrix of Rosenbrock’s function is positive definite at the point x = (1, 1) T . In the following script,<br />
we define the point x and use the rosenbrock function <strong>to</strong> compute the associated Hessian matrix.<br />
Then we use the spec function in order <strong>to</strong> compute the eigenvalues.<br />
39