Convex Optimization

3.18 Adapt the proof of concavity of the log-determinant function in

3.1.5 to show the following. (f) f(x1, x2) = x1 x12, where 0 1, on R2++.
Solution. Concave and quasiconcave. The Hessian is f is not convex or
Proof of first-order convexity condition quasiconvex.
Solution To prove (3.2), we first consider the case n = 1: We show that a
differentiable function f : R R is convex if and only if

,divide both sides by t, get 3.17 Suppose p < 1, p 6= 0. Show that the function
with dom f = Rn++ is concave. This
where we used the eigenvalue decomposition . In includes as special cases
the last equality we express g as a positive weighted sum of convex and taking the limit as t 0 yields (3.4).
functions 1/(1 + ti), hence it is convex. To show sufficiency, assume the function satisfies (3.4) for all x and y in and the
dom f (which is an interval). Choose any x 6= y, and 0 1, and let z = harmonic mean
x + (1 )y. Applying (3.4) twice yields
. Hint. Adapt the proofs for the log-sum-exp
function and the geometric mean in 3.1.5.
Multiplying the first inequality by , the second by 1 , and adding
them yields

which proves that f is convex.

where i, i = 1, . . . , n, are the eigenvalues of . From the

last equality we see that g is a concave function of t on

, since det Z > 0 and the geometric mean is concave on Rn++.

Theory
Now assume that this inequality holds for any x and y, so if ty + (1 t)x
dom f and ty + (1 t)x dom f, we have

Weve seen that g is convex.

Convex functions
Basic properties and examples
A function f : Rn R is convex if dom f is a convex set and if for all x,
y dom f, and with 0 1, we have
Second-order conditions

Geometrically, this inequality means that the line segment between (x,
f(x)) and (y, f(y)), which is the chord from x to y, lies above the graph of
f (figure 3.1). A function f is strictly convex if strict inequality holds in
(3.1) whenever x 6= y and 0 < < 1. We say f is concave if f is convex,
and strictly concave if f is strictly convex.
For an affine function we always have equality in (3.1), so all affine (and
therefore also linear) functions are both convex and concave. 3.14 Convex-concave functions and saddle-points. We say the function
Conversely, any function that is convex and concave is affine. f : Rn Rm R is convex-concave if f(x, z) is a concave function of z, for
Convexity (or concavity) of these examples can be verified in several
A function is convex if and only if it is convex when restricted to any line each fixed x, and a convex function of x, for each fixed z. We also require
ways, such as directly verifying the inequality (3.1), verifying that the
that intersects its domain. In other words f is convex if and only if for all its domain to have the product form dom f = A B, where A Rn and B
Hessian is positive semidefinite, or restricting the function to an
x dom f and Rm are convex.
arbitrary line and verifying convexity of the resulting function of one
variable.

all v, the function g(t) = f(x + tv) is convex (on its domain, {t | x + tv
dom f}).
This property is very useful, since it allows us to check whether a
function is convex by restricting it to a line.
The analysis of convex functions is a well developed field, which we will
not pursue in any depth. One simple result, for example, is that a convex
function is continuous on the relative interior of its domain; it can have
discontinuities only on its relative boundary.
Convex set
3.16 For each of the following functions determine whether it is convex,
concave, quasiconvex, or quasiconcave.
(a) f(x) = ex 1 on R.
Solution. Strictly convex, and therefore quasiconvex. Also quasiconcave
but not concave.
f(x1, x2) = x1x2 on R2++.
Solution. The Hessian of f is which is neither positive semidefinite nor
negative semidefinite. Therefore, f is neither convex nor concave. It is
quasiconcave, since its superlevel sets are convex. It is not quasiconvex.

(c) f(x1, x2) = 1/(x1x2) on R2++.

Solution. The Hessian of f is Therefore, f is convex and quasiconvex. It is
not quasiconcave or concave.

(d) f(x1, x2) = x1/x2 on R2++.

Solution. The Hessian of f is which is not positive or negative
semidefinite. Therefore, f is not convex or concave. It is quasiconvex and
quasiconcave (i.e., quasilinear), since the sublevel and superlevel sets
are halfspaces. 3.9 Second-order conditions for convexity on an affine set. Let F Rn
m, x Rn. The restriction of f : Rn R to the affine set {Fz + x | z
Rm} is defined as the function f : Rm R with

(e) f(x1, x2) = x21/x2 on R R++.

Solution. f is convex, as mentioned on page 72. (See also figure 3.3). This
is easily verified by working out the Hessian: Therefore, f is convex and
quasiconvex. It is not concave or quasiconcave (see the figure).