Volume 3, Number 3, 2017, 367–384
EXAMPLES OF r-CONVEX FUNCTIONS AND CHARACTERIZATIONS OF r-CONVEX FUNCTIONS
ASSOCIATED WITH SECOND-ORDER CONE
CHIEN-HAO HUANG, HONG-LIN HUANG, AND JEIN-SHAN CHEN∗
Abstract. In this paper, we revisit the concept of r-convex functions which were studied in 1970s. We present several novel examples of r-convex functions that are new to the existing literature. In particular, for any given r, we show examples which are r-convex functions. In addition, we extend the concepts of r-convexity and quasi-convexity to the setting associated with second-order cone.
Characterizations about such new functions are established. These generaliza- tions will be useful in dealing with optimization problems involved in second-order cones.
1. Introduction
It is known that the concept of convexity plays a central role in many appli- cations including mathematical economics, engineering, management science, and optimization theory. Moreover, much attention has been paid to its generaliza- tion, to the associated generalization of the results previously developed for the classical convexity, and to the discovery of necessary and/or sufficient conditions for a function to have generalized convexities. Some of the known extensions are quasiconvex functions, r-convex functions [1, 24], and so-called SOC-convex func- tions [7, 8]. Other further extensions can be found in [19, 23]. For a single variable continuous, the midpoint-convex function onR is also a convex function. This result was generalized in [22] by relaxing continuity to lower-semicontinuity and replacing the number 12 with an arbitrary parameter α ∈ (0, 1). An analogous consequence was obtained in [18, 23] for quasiconvex functions.
To understand the main idea behind r-convex function, we recall some concepts that were independently defined by Martos [17] and Avriel [2], and has been studied by the latter author. Indeed, this concept relies on the classical definition of convex functions and some well-known results from analysis dealing with weighted means of positive numbers. Let w = (w1, . . . , wm)∈ Rm, q = (q1, . . . , qm)∈ Rm be vectors
2010 Mathematics Subject Classification. 26A27, 26B05, 90C33.
Key words and phrases. r-Convex function, monotone function, second-order cone, spectral decomposition.
∗Corresponding author. The author’s work is supported by Ministry of Science and Technology, Taiwan.
whose components are positive and nonnegative numbers, respectively, such that
∑m
i=1qi = 1. Given the vector of weights q, the weighted r-mean of the numbers w1, . . . , wm is defined as below (see [13]):
(1.1) Mr(w; q) = Mr(w1, . . . , wm; q) :=
(∑m
i=1
qi(wi)r )1/r
if r ̸= 0,
∏m i=1
(wi)qi if r = 0.
It is well-known from [13] that for s > r, there holds
(1.2) Ms(w1, . . . , wm; q)≥ Mr(w1, . . . , wm; q) for all q1, . . . , qm ≥ 0 with ∑m
i=1qi = 1. The r-convexity is built based on the aforementioned weighted r-mean. For a convex set S ⊆ Rn, a real-valued function f : S ⊆ Rn → R is said to be r-convex if, for any x, y ∈ S, λ ∈ [0, 1], q2 = λ, q1 = 1− q2, q = (q1, q2), there has
f (q1x + q2y)≤ ln{
Mr(ef (x), ef (y); q) }
.
From (1.1), it can be verified that the above inequality is equivalent to (1.3) f ((1− λ)x + λy) ≤
{ ln[(1− λ)erf (x)+ λerf (y)]1/r if r ̸= 0, (1− λ)f(x) + λf(y) if r = 0.
Similarly, f is said to be r-concave on S if the inequality (1.3) is reversed. It is clear from the above definition that a real-valued function is convex (concave) if and only if it is 0-convex (0-concave). Besides, for r < 0 (r > 0), an r-convex (r-concave) function is called superconvex (superconcave); while for r > 0 (r < 0), it is called subconvex (subconcave). In addition, it can be verified that the r-convexity of f on C with r > 0 (r < 0) is equivalent to the convexity (concavity) of erf on S.
A function f : S ⊆ Rn→ R is said to be quasiconvex on S if, for all x, y ∈ S, f (λx + (1− λ)y) ≤ max {f(x), f(y)} , 0 ≤ λ ≤ 1.
Analogously, f is said to be quasiconcave on S if, for all x, y∈ S, f (λx + (1− λ)y) ≥ min {f(x), f(y)} , 0 ≤ λ ≤ 1.
From [13], we know that
r→+∞lim Mr(w1, . . . , wm; q)≡ M∞(w1, . . . , wm) = max{w1, . . . , wm},
r→−∞lim Mr(w1,· · · , wm; q)≡ M−∞(w1, . . . , wm) = min{w1,· · · , wm}.
Then, it follows from (1.2) that M∞(w1, . . . , wm) ≥ Mr(w1, . . . , wm; q) ≥ M−∞(w1, . . . , wm) for every real number r. Thus, if f is r-convex on S, it is also (+∞)-convex, that is, f(λx + (1 − λ)y) ≤ max{f(x), f(y)} for every x, y ∈ S and λ ∈ [0, 1]. Similarly, if f is r-concave on S, it is also (−∞)-concave, i.e., f (λx + (1− λ)y) ≥ min{f(x), f(y)}.
The following review some basic properties regarding r-convex function from [1]
that will be used in the subsequent analysis.
Property 1.1. Let f : S⊆ Rn→ R. Then, the followings hold.
(a) If f is r-convex (r-concave) on S, then f is also s-convex (s-concave) on S for s > r (s < r).
(b) Suppose that f is twice continuously differentiable on S. For any (x, r) ∈ S× R, we define
ϕ(x, r) =∇2f (x) + r∇f(x)∇f(x)T.
Then, f is r-convex on S if and only if ϕ is positive semidefinite for all x∈ S.
(c) Every r-convex (r-concave) function on a convex set S is also quasiconvex (quasiconcave) on S.
(d) f is r-convex if and only if (−f) is (−r)-concave.
(e) Let f be r-convex (r-concave), α∈ R and k > 0. Then f + α is r-convex (r-concave) and k· f is (kr)-convex ((rk)-concave).
(f) Let ϕ, ψ : S ⊆ Rn→ R be r-convex (r-concave) and α1, α2 > 0. Then, the function θ defined by
θ(x) = {
ln[
α1erϕ(x)+ α2erψ(x)]1/r
if r ̸= 0, α1ϕ(x) + α2ψ(x) if r = 0, is also r-convex (r-concave).
(g) Let ϕ : S ⊆ Rn→ R be r-convex (r-concave) such that r ≤ 0 (r ≥ 0) and let the real valued function ψ be nondecreasing s-convex (s-concave) onR with s∈ R. Then, the composite function θ = ψ ◦ ϕ is also s-convex (s-concave).
(h) ϕ : S ⊆ Rn → R is r-convex (r-concave) if and only if, for every x, y ∈ S, the function ψ given by
ψ(λ) = ϕ ((1− λ)x + λy)
is an r-convex (r-concave) function of λ for 0≤ λ ≤ 1.
(i) Let ϕ be a twice continuously differentiable real quasiconvex function on an open convex set S⊆ Rn. If there exists a real number r∗ satisfying
(1.4) r∗ = sup
x∈S, ∥z∥=1
−zT∇2ϕ(x)z [zT∇ϕ(x)]2
whenever zT∇ϕ(x) ̸= 0, then ϕ is r-convex for every r ≥ r∗. We obtain the r-concave analog of the above theorem by replacing supremum in (1.4) by infimum.
In this paper, we will present new examples of r-convex functions in Section 2.
Meanwhile, we extend the r-convexity and quasi-convexity concepts to the setting associated with second-order cone in Section 4 and Section 5. Applications of r- convexity to optimization theory can be found in [2, 12, 15]. In general, r-convex functions can be viewed as the functions between convex functions and quasi-convex
functions. We believe that the aforementioned extensions will be beneficial for deal- ing optimization problems involved second-order constraints. We point out that extending the concepts of r-convex and quasi-convex functions to the setting asso- ciated with second-order cone, which belongs to symmetric cones, is not easy and obvious since any two vectors in the Euclidean Jordan algebra cannot be compared under the partial order≼Kn, see [8]. Nonetheless, using the projection onto second- order cone pave a way to do such extensions, more details will be seen in Sections 4 and 5.
To close this section, we recall some background materials regarding second-order cone. The second-order cone (SOC for short) in Rn, also called the Lorentz cone, is defined by
Kn={
x = (x1, x2)∈ R × Rn−1| ∥x2∥ ≤ x1
}.
For n = 1, Kn denotes the set of nonnegative real number R+. For any x, y in Rn, we write x≽Kn y if x−y ∈ Knand write x≻Kn y if x−y ∈ int(Kn). In other words, we have x ≽Kn 0 if and only if x ∈ Kn and x ≻Kn 0 if and only if x ∈ int(Kn).
The relation ≽Kn is a partial ordering but not a linear ordering in Kn, i.e., there exist x, y ∈ Kn such that neither x≽Kn y nor y ≽Kn x. To see this, for n = 2, let x = (1, 1) and y = (1, 0), we have x− y = (0, 1) /∈ Kn, y− x = (0, −1) /∈ Kn.
For dealing with second-order cone programs (SOCP) and second-order cone com- plementarity problems (SOCCP), we need spectral decomposition associated with SOC [9]. More specifically, for any x = (x1, x2) ∈ R × Rn−1, the vector x can be decomposed as
x = λ1u(1)x + λ2u(2)x ,
where λ1, λ2and u(1)x , u(2)x are the spectral values and the associated spectral vectors of x, respectively, given by
λi = x1+ (−1)i∥x2∥, u(i)x =
{ 1
2(1, (−1)i x∥x22∥) if x2̸= 0,
1
2(1, (−1)iw) if x2= 0.
for i = 1, 2 with w being any vector in Rn−1 satisfying ∥w∥ = 1. If x2 ̸= 0, the decomposition is unique.
For any function f :R → R, the following vector-valued function associated with Kn (n≥ 1) was considered in [7, 8]:
(1.5) fsoc(x) = f (λ1)u(1)x + f (λ2)u(2)x , ∀x = (x1, x2)∈ R × Rn−1.
If f is defined only on a subset ofR, then fsocis defined on the corresponding subset of Rn. The definition (1.5) is unambiguous whether x2̸= 0 or x2 = 0. The cases of fsoc(x) = x1/2, x2, exp(x) are discussed in [10]. In fact, the above definition (1.5) is analogous to the one associated with positive semidefinite cone S+n [20, 21].
Throughout this paper, Rn denotes the space of n-dimensional real column vec- tors, C denotes a convex subset of R, S denotes a convex subset of Rn, and ⟨· , ·⟩
means the Euclidean inner product, whereas∥ · ∥ is the Euclidean norm. The nota- tion “:=” means “define”. For any f :Rn→ R, ∇f(x) denotes the gradient of f at x. C(i)(J ) denotes the family of functions which are defined on J ⊆ Rn toR and have the i-th continuous derivative, whileT means transpose.
2. Examples of r-functions
In this section, we try to discover some new r-convex functions which is verified by applying Property 1.1. With these examples, we have a more complete picture about characterizations of r-convex functions. Moreover, for any given r, we also provide examples which are r-convex functions.
Example 2.1. For any real number p, let f : (0,∞) → R be defined by f(t) = tp. (a) If p > 0, then f is convex for p≥ 1, and (+∞)-convex for 0 < p < 1.
(b) If p < 0, then f is convex.
To see this, we first note that f′(t) = ptp−1, f′′(t) = p(p− 1)tp−2 and sup
s·f′(t)̸=0,|s|=1
−s · f′′(t)· s [s· f′(t)]2 = sup
p̸=0
(1− p)t−p
p =
{ ∞ if 0 < p < 1, 0 if p > 1 or p < 0.
Then, applying Property 1.1 yields the desired result.
Example 2.2. Suppose that f is defined on (−π2,π2).
(a) The function f (t) = sin t is∞-convex.
(b) The function f (t) = tan t is 1-convex.
(c) The function f (t) = ln(sec t) is (−1)-convex.
(d) The function f (t) = ln|sec t + tan t| is 1-convex.
To see (a), we note that f′(t) = cos t, f′′(t) =− sin t, and
−π2<t<supπ2,|s|=1
−s · f′′(t)· s
[s· f′(t)]2 = sup
−π2<t<π2
sin t cos2t =∞.
Hence f (t) = sin t is∞-convex.
To see (b), we note that f′(t) = sec2t, f′′(t) = 2 sec2t· tan t, and sup
−π2<t<π2
−f′′(t)
[f′(t)]2 = sup
−π2<t<π2
−2 sec2t· tan t
sec4t = sup
−π2<t<π2
(− sin 2t) = 1.
This says that f (t) = tan t is 1-convex.
To see (c), we note that f′(t) = tan t, f′′(t) = sec2t, and sup
−π2<t<π2
−f′′(t)
[f′(t)]2 = sup
−π2<t<π2
−k sec2t
tan2t = sup
−π2<t<π2
(− csc2t) =−1.
Then, it is clear to see that f (t) = ln(sec t) is (−1)-convex.
To see (d), we note that f′(t) = sec t, f′′(t) = sec t· tan t, and sup
−π2<t<π2
−f′′(t)
[f′(t)]2 = sup
−π2<t<π2
− sec t · tan t
sec2t = sup
−π2<t<π2
(− sin t) = 1.
Thus, f (t) = ln|sec t + tan t| is 1-convex.
Figure 1. Graphs of r-convex functions with various values of r.
In light of Example 2.2(b)-(c) and Property 1.1(e), the next example indicates that for any given r∈ R (no matter positive or negative), we can always construct an r-convex function accordingly. The graphs of various r-convex functions are depicted in Figure 1.
Example 2.3. For any r̸= 0, let f be defined on (−π2,π2).
(a) The function f (t) = tan t
r is|r|-convex.
(b) The function f (t) = ln(sec t)
r is (−r)-convex.
(a) First, we compute that f′(t) = sec2t
r , f′′(t) = 2 sec2t· tan t
r , and
−π2sup<t<π2
−f′′(t)
[f′(t)]2 = sup
−π2<t<π2
(−r sin 2t) = |r|.
This says that f (t) = tan t
r is |r|-convex.
(b) Similarly, from f′(t) = tan t
r , f′′(t) = sec2t r , and sup
−π2<t<π2
−f′′(t)
[f′(t)]2 = sup
−π2<t<π2
(−r csc2t) =−r.
Then, it is easy to see that f (t) = ln(sec t)
r is (−r)-convex.
Example 2.4. The function f (x) = 12ln(∥x∥2+ 1) defined onR2 is 1-convex.
For x = (s, t)∈ R2, and any real number r̸= 0, we consider the function ϕ(x, r) = ∇2f (x) + r∇f(x)∇f(x)T
= 1
(∥x∥2+ 1)2
[t2− s2+ 1 −2st
−2st s2− t2+ 1 ]
+ r
(∥x∥2+ 1)2
[s2 st st t2 ]
= 1
(∥x∥2+ 1)2
[(r− 1)s2+ t2+ 1 (r− 2)st (r− 2)st s2+ (r− 1)t2+ 1
] .
Applying Property 1.1(b), we know that f is r-convex if and only if ϕ is positive semidefinite, which is equivalent to
(r− 1)s2+ t2+ 1≥ 0 (2.1)
(r− 1)s2+ t2+ 1 (r− 2)st (r− 2)st s2+ (r− 1)t2+ 1
≥ 0.
(2.2)
It is easy to verify the inequality (2.1) holds for all x ∈ R2 if and only if r ≥ 1.
Moreover, we note that
(r− 1)s2+ t2+ 1 (r− 2)st (r− 2)st s2+ (r− 1)t2+ 1
≥ 0
⇐⇒ s2t2+ s2+ t2+ 1 + (r− 1)2s2t2+ (r− 1)(s4+ s2+ t4+ t2)− (r − 2)2s2t2≥ 0
⇐⇒ s2+ t2+ 1 + (2r− 2)s2t2+ (r− 1)(s4+ s2+ t4+ t2)≥ 0,
and hence the inequality (2.2) holds for all x ∈ R2 whenever r ≥ 1. Thus, we conclude by Property 1.1(b) that f is 1-convex onR2.
Figure 2. Graphs of 1-convex functions f (x) = 12ln(∥x∥2+ 1).
3. Properties of SOC-functions
As mentioned in Section 1, another contribution of this paper is extending the concept of r-convexity to the setting associated with second-order cone. To this end, we recall what SOC-convex function means. For any x = (x1, x2)∈ R × Rn−1 and y = (y1, y2)∈ R × Rn−1, we define their Jordan product as
x◦ y = (xTy , y1x2+ x1y2).
We write x2 to mean x◦ x and write x + y to mean the usual componentwise addition of vectors. Then, ◦, +, together with e′ = (1, 0, . . . , 0)T ∈ Rn and for any x, y, z ∈ Rn, the following basic properties [10, 11] hold: (1) e′ ◦ x = x, (2) x◦ y = y ◦ x, (3) x ◦ (x2◦ y) = x2◦ (x ◦ y), (4) (x + y) ◦ z = x ◦ z + y ◦ z. Notice that the Jordan product is not associative in general. However, it is power associative, i.e., x◦ (x ◦ x) = (x ◦ x) ◦ x for all x ∈ Rn. Thus, we may, without loss of ambiguity, write xm for the product of m copies of x and xm+n = xm ◦ xn for all positive integers m and n. Here, we set x0 = e′. Besides, Kn is not closed under Jordan product.
For any x∈ Kn, it is known that there exists a unique vector in Kn denoted by x1/2 such that (x1/2)2 = x1/2◦ x1/2= x. Indeed,
x1/2 = (
s,x2 2s
)
, where s =
√ 1 2
( x1+
√
x21− ∥x2∥2 )
.
In the above formula, the term x2/s is defined to be the zero vector if x2 = 0 and s = 0, i.e., x = 0. For any x∈ Rn, we always have x2 ∈ Kn, i.e., x2 ≽Kn 0. Hence, there exists a unique vector (x2)1/2 ∈ Kn denoted by |x|. It is easy to verify that
|x| ≽Kn 0 and x2 =|x|2 for any x ∈ Rn. It is also known that|x| ≽Kn x. For any x ∈ Rn, we define [x]+ to be the nearest point projection of x onto Kn, which is the same definition as in Rn+. In other words, [x]+ is the optimal solution of the parametric SOCP: [x]+= arg min{∥x − y∥ | y ∈ Kn}. In addition, it can be verified that [x]+= (x +|x|)/2; see [10, 11].
Property 3.1 ([11, Proposition 3.3]). For any x = (x1, x2)∈ R × Rn−1, we have (a) |x| = (x2)1/2=|λ1|u(1)x +|λ2|u(2)x .
(b) [x]+= [λ1]+u(1)x + [λ2]+u(2)x = 12(x +|x|).
Next, we review the concepts of SOC-monotone and SOC-convex functions which are introduced in [7].
Definition 3.2. For a real valued function f :R → R,
(a) f is said to be SOC-monotone of order n if its corresponding vector-valued function fsoc defined as in (1.5) satisfies
x≽Kn y =⇒ fsoc(x)≽Kn fsoc(y).
The function f is said to be SOC-monotone if f is SOC-monotone of all order n.
(b) f is said to be SOC-convex of order n if its corresponding vector-valued function fsoc defined as in (1.5) satisfies
(3.1) fsoc((1− λ)x + λy) ≼Kn (1− λ)fsoc(x) + λfsoc(y)
for all x, y ∈ Rn and 0 ≤ λ ≤ 1. Similarly, f is said to be SOC-concave of order n on C if the inequality (3.1) is reversed. The function f is said to be SOC-convex (respectively, SOC-concave) if f is SOC-convex of all order n (respectively, SOC-concave of all order n).
The concepts of SOC-monotone and SOC-convex functions are analogous to ma- trix monotone and matrix convex functions [5, 14], and are special cases of operator monotone and operator convex functions [3, 6, 16]. Examples of SOC-monotone and SOC-convex functions are given in [7]. It is clear that the set of SOC-monotone functions and the set of SOC-convex functions are both closed under linear combi- nations and under pointwise limits.
Property 3.3 ([8, Theorem 3.1]). Let f ∈ C(1)(J ) with J being an open interval and dom(fsoc)⊆ Rn. Then, the following hold.
(a) f is SOC-monotone of order 2 if and only if f′(τ )≥ 0 for any τ ∈ J;
(b) f is SOC-monotone of order n≥ 3 if and only if the 2 × 2 matrix
f(1)(t1) f (t2)− f(t1) t2− t1
f (t2)− f(t1) t2− t1
f(1)(t2)
≽ O for all t1, t2 ∈ J and t1 ̸= t2.
Property 3.4 ([8, Theorem 4.1]). Let f ∈ C(2)(J ) with J being an open interval inR and dom(fsoc)⊆ Rn. Then, the following hold.
(a) f is SOC-convex of order 2 if and only if f is convex;
(b) f is SOC-convex of order n≥ 3 if and only if f is convex and the inequality (3.2) 1
2f(2)(t0)[f (t0)− f(t) − f(1)(t)(t0− t)]
(t0− t)2 ≥ [f (t)− f(t0)− f(1)(t0)(t− t0)]
(t0− t)4 holds for any t0, t∈ J and t0 ̸= t.
Property 3.5 ([4, Theorem 3.3.7]). Let f : S → R where S is a nonempty open convex set inRn. Suppose f ∈ C2(S). Then, f is convex if and only if∇2f (x)≽ O, for all x∈ S.
Property 3.6 ([7, Proposition 4.1]). Let f : [0,∞] → [0, ∞] be continuous. If f is SOC-concave, then f is SOC-monotone.
Property 3.7 ( [11, Proposition 3.2]). Suppose that f (t) = et and g(t) = ln t.
Then, the corresponding SOC-functions of et and ln t are given as below.
(a) For any x = (x1, x2)∈ R × Rn−1, fsoc(x) = ex=
{ ex1
(cosh(∥x2∥), sinh(∥x2∥)∥xx22∥)
if x2 ̸= 0,
(ex1, 0) if x2 = 0,
where cosh(α) = (eα+ e−α)/2 and sinh(α) = (eα− e−α)/2 for α∈ R.
(b) For any x = (x1, x2)∈ int(Kn), ln x is well-defined and
gsoc(x) = ln x = { 1
2
(
ln(x21− ∥x2∥2), ln
(x1+∥x2∥ x1−∥x2∥
) x2
∥x2∥
)
if x2̸= 0,
(ln x1, 0) if x2= 0.
With these, we have the following technical lemmas that will be used in the subsequent analysis.
Lemma 3.8. Let f : R → R be f(t) = et and x = (x1, x2) ∈ R × Rn−1, y = (y1, y2)∈ R × Rn−1. Then, the following hold.
(a) f is SOC-monotone of order 2 on R.
(b) f is not SOC-monotone of order n≥ 3 on R.
(c) If x1− y1 ≥ ∥x2∥ + ∥y2∥, then ex ≽Kn ey. In particular, if x ∈ Kn, then ex ≽Kn e(0,0).
Proof. (a) By applying Property 3.3(a), it is clear that f is SOC-monotone of order 2 since f′(τ ) = eτ ≥ 0 for all τ ∈ R.
(b) Take x = (2, 1.2,−1.6), y = (−1, 0, −4), then we have x−y = (3, 1.2, 2.4) ≽Kn 0.
But, we compute that
ex− ey = e2 (
cosh(2), sinh(2)(1.2,−1.6) 2
)
− e−1 (
cosh(4), sinh(4)(0,−4) 4
)
= 1
2
[(e4+ 1, .6(e4− 1), −.8(e4− 1)) − (e3+ e−5, 0,−e3+ e−5)]
= (17.7529, 16.0794,−11.3999) Kn 0.
The last inequality is because ∥(16.0794, −11.3999)∥ = 19.7105 > 17.7529.
We also present an alternative argument for part(b) here. First, we observe that
(3.3) det
[ f(1)(t1) f (t2t)−f(t1)
2−t1
f (t2)−f(t1)
t2−t1 f(1)(t2) ]
= et1+t2 −
(et2 − et1 t2− t1
)2
≥ 0
if and only if 1≥ (
e(t2−t1)/2− e(t1−t2)/2 t2− t1
)2
. Denote s := (t2−t1)/2, then the above inequality holds if and only if 1 ≥ (sinh(s)/s)2. In light of Taylor Theorem, we know sinh(s)/s = 1 + s2/6 + s4/120 +· · · > 1 for s ̸= 0. Hence, (3.3) does not hold.
Then, applying Property 3.3(b) says f is not SOC-monotone of order n≥ 3 on R.
(c) The desired result follows by the following implication:
ex ≽Kn ey
⇐⇒ ex1cosh(∥x2∥) − ey1cosh(∥y2∥)
≥
ex1sinh(∥x2∥) x2
∥x2∥ − ey1sinh(∥y2∥) y2
∥y2∥
⇐⇒ [ex1cosh(∥x2∥) − ey1cosh(∥y2∥)]2
−
ex1sinh(∥x2∥) x2
∥x2∥− ey1sinh(∥y2∥) y2
∥y2∥ 2
= e2x1 + e2y1
−2ex1+y1 [
cosh(∥x2∥) cosh(∥y2∥) − sinh(∥x2∥) sinh(∥y2∥) ⟨x2, y2⟩
∥x2∥∥y2∥ ]
≥ 0
⇐= e2x1+ e2y1 − 2ex1+y1cosh(∥x2∥ + ∥y2∥) ≥ 0
⇐⇒ cosh(∥x2∥ + ∥y2∥) ≤ e2x1 + e2y1
2ex1+y1 = ex1−y1 + ey1−x1
2 = cosh(x1− y1)
⇐⇒ x1− y1 ≥ ∥x2∥ + ∥y2∥.
Lemma 3.9. Let f (t) = et be defined on R, then f is SOC-convex of order 2.
However, f is not SOC-convex of order n≥ 3.
Proof. (a) By applying Property 3.4 (a), it is clear that f is SOC-convex since exponential function is a convex function onR.
(b) As below, it is a counterexample which shows f (t) = et is not SOC-convex of order n≥ 3. To see this, we compute that
e[(2,0,−1)+(6,−4,−3)]/2 = e(4,−2,−2)
= e4 (
cosh(2√
2) , sinh(2√
2)· (−2, −2)/(2√ 2)
) + (463.48, −325.45, −325.45)
and
1 2
(
e(2,0,−1)+ e(6,−4,−3) )
= 1
2
[e2(cosh(1), 0,− sinh(1)) + e6(cosh(5), sinh(5)· (−4, −3)/5)]
= (14975,−11974, −8985).
We see that 14975− 463.48 = 14511.52, but
∥(−11974, −8985) − (−325.4493, −325.4493)∥ = 14515 > 14511.52 which is a contradiction.
Lemma 3.10 ([8, Proposition 5.1]). The function g(t) = ln t is SOC-monotone of order n≥ 2 on (0, ∞).
In general, to verify the SOC-convexity of et (as shown in Proposition 3.1), we observe that the following fact
0≺Kn erfsoc(λx+(1−λ)y)≼Knw =⇒ rfsoc(λx + (1− λ)y) ≼Kn ln(w) is important and often needed. Note for x2 ̸= 0, we also have some observations as below.
(a) ex ≻Kn 0 ⇐⇒ cosh(∥x2∥) ≥ | sinh(∥x2∥)| ⇐⇒ e−∥x2∥ > 0 . (b) 0 ≺Kn ln(x) ⇐⇒ ln(x21− ∥x2∥2) > ln(
x1+∥x2∥
x1−∥x2∥) ⇐⇒ ln(x1 − ∥x2∥) >
0 ⇐⇒ x1− ∥x2∥ > 1. Hence (1, 0) ≺Kn x implies 0≺Kn ln(x).
(c) ln(1, 0) = (0, 0) and e(0,0)= (1, 0).
4. SOC-r-convex functions
In this section, we define the so-called SOC-r-convex functions which is viewed as the natural extension of r-convex functions to the setting associated with second- order cone.
Definition 4.1. Suppose that r ∈ R and f : C ⊆ R → R where C is a convex subset of R. Let fsoc : S ⊆ Rn→ Rnbe its corresponding SOC-function defined as in (1.5). The function f is said to be SOC-r-convex of order n on C if, for x, y∈ S and λ∈ [0, 1], there holds
(4.1) fsoc(λx + (1− λ)y) ≼Kn {1
rln(
λerfsoc(x)+ (1− λ)erfsoc(y))
r ̸= 0, λfsoc(x) + (1− λ)fsoc(y) r = 0.
Similarly, f is said to be SOC-r-concave of order n on C if the inequality (4.1) is reversed. We say f is SOC-r-convex (respectively, SOC-r-concave) on C if f is SOC-r-convex of all order n (respectively, SOC-r-concave of all order n) on C.
It is clear from the above definition that a real function is SOC-convex (SOC- concave) if and only if it is SOC-0-convex (SOC-0-concave). In addition, a function f is SOC-r-convex if and only if−f is SOC-(−r)-concave. From [1, Theorem 4.1], it is shown that ϕ :R → R is r-convex with r ̸= 0 if and only if erϕis convex whenever r > 0 and concave whenever r < 0. However, we observe that the exponential function et is not SOC-convex for n≥ 3 by Lemma 3.9. This is a hurdle to build parallel result for general n in the setting of SOC case. As seen in Proposition 4.5, the parallel result is true only for n = 2. Indeed, for n≥ 3, only one direction holds which can be viewed as a weaker version of [1, Theorem 4.1].
Proposition 4.2. Let f : [0,∞) → [0, ∞) be continuous. If f is SOC-r-concave with r≥ 0, then f is SOC-monotone.
Proof. For any 0 < λ < 1, we can write λx = λy +(1(1−λ)λ−λ) (x− y). If r = 0, then f is SOC-concave and SOC-monotone by Property 3.6. If r > 0, then
fsoc(λx) ≽Kn 1 r ln
(
λerfsoc(y)+ (1− λ)erfsoc(1−λλ (x−y)))
≽Kn 1 r ln
(
λer(0,0)+ (1− λ)er(0,0))
= 1
r ln (λ(1, 0) + (1− λ)(1, 0))
= 0,
where the second inequality is due to x− y ≽Kn 0 and Lemmas 3.8-3.10. Letting λ→ 1, we obtain that fsoc(x) ≽Kn fsoc(y), which says that f is SOC-monotone.
In fact, in light of Lemma 3.8-3.10, we have the following Lemma which is useful for subsequent analysis.
Lemma 4.3. Let z∈ Rn and w∈ int(Kn). Then, the following hold.
(a) For n = 2 and r > 0, z≼Kn ln(w)/r⇐⇒ rz ≼Kn ln(w)⇐⇒ erz ≼Kn w.
(b) For n = 2 and r > 0, z≼Kn ln(w)/r⇐⇒ rz ≽Kn ln(w)⇐⇒ erz ≽Kn w.
(c) For n≥ 2, if erz ≼Kn w, then rz≼Kn ln(w).
Proposition 4.4. For n = 2 and let f :R → R. Then, the following hold.
(a) The function f (t) = t is SOC-r-convex (SOC-r-concave) on R for r > 0 (r < 0).
(b) If f is SOC-convex, then f is SOC-r-convex (SOC-r-concave) for r > 0 (r < 0).
Proof. (a) For r > 0, x, y ∈ Rn and λ ∈ [0, 1], we note that the corresponding vector-valued SOC-function of f (t) = t is fsoc(x) = x. Therefore, to prove the desired result, we need to verify that
fsoc(λx + (1− λ)y) ≼Kn 1 rln
(
λerfsoc(x)+ (1− λ)erfsoc(y)) . To this end, we see that
λx + (1− λ)y ≼Kn 1
r ln (λerx+ (1− λ)ery)
⇐⇒ λrx + (1 − λ)ry ≼Kn ln (λerx+ (1− λ)ery)
⇐⇒ eλrx+(1−λ)ry≼Kn λerx+ (1− λ)ery,
where the first “⇐⇒” is true due to Lemma 4.3, whereas the second “⇐⇒” holds because et and ln t are SOC-monotone of order 2 by Lemma 3.8 and Lemma 3.10.
Then, using the fact that et is SOC-convex of order 2 gives the desired result.
(b) For any x, y∈ Rn and 0≤ λ ≤ 1, it can be verified that fsoc(λx + (1− λ)y) ≼Kn λfsoc(x) + (1− λ)fsoc(y)
≼Kn 1 rln
(
λerfsoc(x)+ (1− λ)erfsoc(y)) ,
where the second inequality holds according to the proof of (a). Thus, the desired result follows.
Proposition 4.5. Let f : R → R. Then f is SOC-r-convex if erf is SOC-convex (SOC-concave) for n≥ 2 and r > 0 (r < 0). For n = 2, we can replace “if” by “if and only if”.
Proof. Suppose that erf is SOC-convex. For any x, y ∈ Rn and 0≤ λ ≤ 1, using that fact that ln t is SOC-monotone (Lemma 3.10) yields
erfsoc(λx+(1−λ)y)≼Knλerfsoc(x)+ (1− λ)erfsoc(y)
=⇒ rfsoc(λx + (1− λ)y) ≼Kn ln (
λerfsoc(x)+ (1− λ)erfsoc(y))
⇐⇒ fsoc(λx + (1− λ)y) ≼Kn 1 rln
(
λerfsoc(x)+ (1− λ)erfsoc(y)) .
When n = 2, et is SOC-monotone as well, which implies that the “=⇒” can be replaced by “⇐⇒”. Thus, the proof is complete.
Combining with Property 3.4, we can characterize the SOC-r-convexity as follows.
Proposition 4.6. Let f ∈ C(2)(J ) with J being an open interval inR and dom(fsoc)⊆ Rn. Then, for r > 0, the followings hold.
(a) f is SOC-r-convex of order 2 if and only if erf is convex;
(b) f is SOC-r-convex of order n≥ 3 if erf is convex and satisfies the inequality (3.2).
Next, we present several examples of SOC-r-convex and SOC-r-concave functions of order 2. For examples of SOC-r-convex and SOC-r-concave functions (of order n), we are still unable to discover them.
Example 4.7. For n = 2, the following hold.
(a) The function f (t) = t2 is SOC-r-convex onR for r ≥ 0.
(b) The function f (t) = t3 is SOC-r-convex on [0,∞) for r > 0, while it is SOC-r-concave on (−∞, 0] for r < 0.
(c) The function f (t) = 1/t is SOC-r-convex on [−r/2, 0) or (0, ∞) for r > 0, while it is SOC-r-concave on (−∞, 0) or (0, −r/2] for r < 0.
(d) The function f (t) =√
t is SOC-r-convex on [1/r2,∞) for r > 0, while it is SOC-r-concave on [0,∞) for r < 0.
(e) The function f (t) = ln t is SOC-r-convex (SOC-r-concave) on (0,∞) for r > 0 (r < 0).
Proof. (a) First, we denote h(t) := ert2. Then, we have h′(t) = 2rtert2 and h′′(t) = (1 + 2rt2)2rert2. From Property 3.5, we know h is convex if and only if h′′(t)≥ 0.
Thus, the desired result holds by applying Property 3.4 and Proposition 4.5. The arguments for other cases are similar and we omit them.
5. SOC-quasiconvex functions
In this section, we define the so-called SOC-quasiconvex functions which is a natural extension of quasiconvex functions to the setting associated with second- order cone.
Recall that a function f : S ⊆ Rn→ R is said to be quasiconvex on S if, for any x, y∈ S and 0 ≤ λ ≤ 1, there has
f (λx + (1− λ)y) ≤ max {f(x), f(y)} .
We point out that the relation≽Kn is not a linear ordering. Hence, it is not possible to compare any two vectors (elements) via≽Kn. Nonetheless, we note that
max{a, b} = b + [a − b]+= 1
2(a + b +|a − b|), for any a, b ∈ R.
This motivates us to define SOC-quasiconvex functions in the setting of second-order cone.
Definition 5.1. Let f : C ⊆ R → R and 0 ≤ λ ≤ 1. The function f is said to be SOC-quasiconvex of order n on C if, for any x, y∈ Rn, there has
fsoc(λx + (1− λ)y) ≼Kn fsoc(y) + [fsoc(x)− fsoc(y)]+ where
fsoc(y) + [fsoc(x)− fsoc(y)]+
=
fsoc(x) if fsoc(x)≽Kn fsoc(y), fsoc(y) if fsoc(x)≺Kn fsoc(y),
1
2(fsoc(x) + fsoc(y) +|fsoc(x)− fsoc(y)|) if fsoc(x)− fsoc(y) /∈ Kn∪ (−Kn).
Similarly, f is said to be SOC-quasiconcave of order n if
fsoc(λx + (1− λ)y) ≽Kn fsoc(x)− [fsoc(x)− fsoc(y)]+.
The function f is called SOC-quasiconvex (SOC-quasiconcave) if it is SOC-quasiconvex of all order n (SOC-quasiconcave of all order n).
Proposition 5.2. Let f :R → R be f(t) = t. Then, f is SOC-quasiconvex on R.
Proof. First, for any x = (x1, x2) ∈ R × Rn−1, y = (y1, y2) ∈ R × Rn−1, and 0≤ λ ≤ 1, we have
fsoc(y)≼Knfsoc(x) ⇐⇒ (1 − λ)fsoc(y)≼Kn (1− λ)fsoc(x)
⇐⇒ λfsoc(x) + (1− λ)fsoc(y)≼Knfsoc(x).
Recall that the corresponding SOC-function of f (t) = t is fsoc(x) = x. Thus, for all x∈ Rn, this implies fsoc(λx+(1−λ)y) = λfsoc(x)+(1−λ)fsoc(y)≼Kn fsoc(x) under this case: fsoc(y)≼Kn fsoc(x). The argument is similar to the case of fsoc(x)≼Kn
fsoc(y). Hence, it remains to consider the case of fsoc(x)− fsoc(y) /∈ Kn∪ (−Kn), i.e., it suffices to show that
λfsoc(x) + (1− λ)fsoc(y)≼Kn 1
2(fsoc(x) + fsoc(y) +|fsoc(x)− fsoc(y)|) .
To this end, we note that
|fsoc(x)− fsoc(y)| ≽Knfsoc(x)− fsoc(y) and
|fsoc(x)− fsoc(y)| ≽Kn fsoc(y)− fsoc(x), which respectively implies
1
2(fsoc(x) + fsoc(y) +|fsoc(x)− fsoc(y)|) ≽Kn x, (5.1)
1
2(fsoc(x) + fsoc(y) +|fsoc(x)− fsoc(y)|) ≽Kn y.
(5.2)
Then, adding up (5.1) ×λ and (5.2) ×(1 − λ) yields the desired result. Proposition 5.3. If f : C ⊆ R → R is SOC-convex on C, then f is also SOC- quasiconvex on C.
Proof. For any x, y ∈ Rn and 0≤ λ ≤ 1, it can be verified that
fsoc(λx + (1−λ)y) ≼Kn λfsoc(x) + (1−λ)fsoc(y)≼Kn fsoc(y) + [fsoc(x)− fsoc(y)]+, where the second inequality holds according to the proof of Proposition 5.2. Thus, the desired result follows.
From Proposition 5.3, we can easily construct examples of SOC-quasiconvex func- tions. More specifically, all the SOC-convex functions which were verified in [7] are SOC-quasiconvex functions, for instances, t2 on R, and t3, 1t, t1/2 on (0,∞).
6. Final remarks
In this paper, we revisit the concept of r-convex functions and provide a way to construct r-convex functions for any given r∈ R. We also extend such concept to the setting associated with SOC which will be helpful in dealing with optimization problems involved in second-order cones. In particular, we obtain some characteri- zations for SOC-r-convexity and SOC-quasiconvexity.
Indeed, this is just the first step and there still have many things to clarify. For example, in Section 4, we conclude that SOC-convexity implies SOC-r-convexity for n = 2 only. The key role therein relies particularly on the SOC-convexity and SOC- monotonicity of et. However, for n > 2, the expressions of ex and ln(x) associated with second-order cone are very complicated so that it is hard to compare any two elements. In other words, when n = 2, the SOC-convexity and SOC-monotonicity of etmake things much easier than the general case n≥ 3. To conquer this difficulty, we believe that we have to derive more properties of ex. In particular, “Does SOC- r-convex function have similar results as shown in Property 1.1?” is an important future direction.
Manuscript received revised