We study the notions of strongly convex function as well as F-strongly convex function. We present here some new integral inequalities of Jensen’s type for these classes of functions. A refinement of companion inequality to Jensen’s inequality established by Matić and Pečarić is shown to be recaptured as a particular instance. Counterpart of the integral Jensen inequality for strongly convex functions is also presented. Furthermore, we present integral Jensen-Steffensen and Slater’s inequality for strongly convex functions.
National Natural Science Foundation of China61673169116014851. Introduction and Preliminaries
The word “convexity” is the most important, natural, and fundamental notations in mathematics. Convex functions were presented by Johan Jensen over 100 years ago. Over the past few years, multiple generalizations and extensions have been made for convexity. These extensions and generalizations in the theory of inequalities have made valuable contributions in many areas of mathematics. Some new generalized concepts in this point of view are quasiconvex [1], strongly convex [2], approximately convex [3], logarithmically convex [4], midconvex functions [5], pseudoconvex [6], φ-convex [7], λ-convex [8], h-convex [9], delta-convex [10], Schur convex [11–15], and others [16–19].
The main ingredient of our investigation is the strongly convex function [2]. Let Ψ be the real function defined on interval I and c be positive number, then we say that the function Ψ is strongly convex with modulus c on I if(1)Ψηz+1-ηy≤ηΨz+1-ηΨy-cη1-ηz-y2for all y,z∈I and η∈[0,1].
Every strongly convex function is convex, but the converse is not true in general. Strongly convex functions have been utilized for proving the convergence of a gradient type algorithm for minimizing a function. They play a significant role in mathematical economics, approximation theory, and optimization theory. Many applications and properties of them can be found in [2, 9, 20]. In 2016, Adamek [21] further generalized the notion of strongly convex function. They replaced the nonnegative term cz-y2 by a nonnegative real valued function F and defined it as follows: function Ψ is said to be F-strongly convex function if(2)Ψηz+1-ηy≤ηΨz+1-ηΨy-η1-ηFz-yfor all y,z∈I and η∈0,1. From [22], we also have(3)Ψy+Ψ+′yz-y+Fz-y≤Ψz,where Ψ is F-strongly convex function.
In literature the following inequality is well-known as Jensen inequality.
Theorem 1 (see [4]).
Let (Λ,X,μ) be a measure space with 0<μ(Λ)<∞ and Ψ:I→R be convex function. Suppose ϕ:Λ→I is such that ϕ, Ψ(ϕ)∈L1(μ), then one has(4)Ψ1μΛ∫Λϕdμ≤1μΛ∫ΛΨϕdμ.
In 1981, Slater proved a companion inequality to the Jensen inequality [23].
Theorem 2.
Let (Λ,X,μ) be a measure space with 0<μ(Λ)<∞ and Ψ:I→R be increasing and convex function. Suppose ϕ:Λ→I is such that Ψ(ϕ), Ψ+′(ϕ), and ϕΨ+′(ϕ)∈L1(μ). If ∫ΛΨ+′(ϕ)dμ≠0, then one has(5)1μΛ∫ΛΨϕdμ≤Ψ∫ΛϕΨ+′ϕdμ∫ΛΨ+′ϕdμ.In the case when Ψ is strictly convex, then equality holds in (5) if and only if ϕ is constant almost everywhere on Λ.
Remark 3.
Some improvements and reversions of Slater’s inequality are given in [24, 25].
The following inequality is the integral analogue of another companion inequality to the Jensen inequality.
Theorem 4 (see [26]).
Let (Λ,X,μ) be a measure space with 0<μ(Λ)<∞ and Ψ:I→R be convex function. Suppose ϕ:Λ→I is such that ϕ, Ψ(ϕ), Ψ+′(ϕ), and ϕΨ+′(ϕ)∈L1(μ) and (6)ϕ¯=1μΛ∫ΛϕdμΨ¯=1μΛ∫ΛΨϕdμ.Then the following inequalities hold:(7)0≤Ψ¯-Ψϕ¯≤1μΛ∫Λϕ-ϕ¯Ψ+′ϕdμ.If the function Ψ is strictly convex, then equality holds in (7) if and only if ϕ is constant almost everywhere on Λ.
Matić and Pečarić established a general inequality from which one can directly obtain inequalities (5) and (7).
Theorem 5 (see [27]).
Let all the assumptions of Theorem 4 be fulfilled. If d1,d2∈I, then one has(8)Ψd1+Ψ+′d1ϕ¯-d1≤Ψ¯≤Ψd2+1μΛ∫Λϕ-d2Ψ+′ϕdμ.Also, when Ψ is strictly convex, then equality in the left side in (8) holds if and only if ϕ=d1 almost everywhere on Λ, while equality in the right side in (8) holds if and only if ϕ=d2 almost everywhere on Λ.
Remark 6.
Under the assumptions of Theorem 4, let ∫ΛΨ+′(ϕ)dμ≠0 and Ψ¯¯=∫ΛϕΨ+′(ϕ)dμ/∫ΛΨ+′(ϕ)dμ∈I, then by setting d2=Ψ¯¯ in (8), we get Slater’s inequality (5), and similarly by setting d2=ϕ¯ in (8), we get (7).
Merentes and Nikodem improved the Jensen inequality for strongly convex functions as follows.
Theorem 7 (see [28]).
Let (Λ,X,μ) be a probability measure space and Ψ:I→R be strongly convex function with modulus c. Suppose ϕ:Λ→I is a Lebesgue integrable function and θ¯=∫Λϕdμ. Then the following inequality holds:(9)0≤∫ΛΨϕdμ-Ψθ¯-c∫Λϕ-θ¯2dμ.
For more recent results related to strongly convex function and Jensen type inequalities we recommend [22, 29–34].
This paper is organized as follows. In Section 2, we establish general inequalities for F-strongly convex function as well as strongly convex functions. As a consequence, we obtain integral Jensen inequality and Slater’s inequality for strongly convex functions. Also by the virtue of these general inequalities we deduce converse of Jensen inequality. In Section 3, we give some properties of strongly convex functions. By using these properties of strongly convex functions we prove Jensen-Steffensen and Slater’s type inequalities.
2. Jensen’s Type Inequalities
We start this section to give the following general theorem.
Theorem 8.
Let (Λ,X,μ) be a measure space with 0<μ(Λ)<∞ and Ψ:I→R be F-strongly convex function. Suppose ϕ:Λ→I is such that , Ψ(ϕ), Ψ+′(ϕ), and ϕΨ+′(ϕ)∈L1(μ) and also ϕ¯=1/μ(Λ)∫Λϕdμ, Ψ¯=1/μ(Λ)∫ΛΨ(ϕ)dμ. If d1,d2∈I, then one has(10)Ψd1+Ψ+′d1ϕ¯-d1+1μΛ∫ΛFϕ-d1dμ≤Ψ¯≤Ψd2+1μΛ∫Λϕ-d2Ψ+′ϕdμ-1μΛ∫ΛFd2-ϕdμ.
Proof.
Since Ψ is strongly convex function, therefore(11)Ψy-Ψ+′yy-z+Fz-y≤Ψz.Letting y→ϕ and z→d2, in (11), we get(12)Ψϕ-Ψ+′ϕϕ-d2+Fd2-ϕ≤Ψd2.Taking integral of (12) and then dividing by μ(Λ), we obtain(13)Ψ¯≤Ψd2+1μΛ∫Λϕ-d2Ψ+′ϕdμ-1μΛ∫ΛFd2-ϕdμ.Similarly, rearranging (11), we get(14)Ψy+Ψ+′yz-y+Fz-y≤Ψz.Letting y→d1 and z→ϕ, in (14), we have(15)Ψd1+Ψ+′d1ϕ-d1+Fϕ-d1≤Ψϕ.Taking integral of (15) and then dividing by μ(Λ), we get(16)Ψd1+Ψ+′d1ϕ¯-d1+1μΛ∫ΛFϕ-d1dμ≤Ψ¯.Combining (13) and (16), we obtain (10).
By virtue of Theorem 8, we can deduce some new and interesting consequences.
Proposition 9.
Suppose that all the assumptions of Theorem 8 are satisfied. Then(17)1μΛ∫ΛΨϕdμ≤1μΛ∫ΛϕΨ+′ϕdμ+infz∈IΨz-zμΛ∫ΛΨ+′ϕdμ-infz∈I1μΛ∫ΛFz-ϕdμ.
Proof.
If we set y=ϕ in (11) and taking integral over Λ and then dividing by μ(Λ), we have(18)1μΛ∫ΛΨϕdμ-1μΛ∫ΛϕΨ+′ϕdμ+zμΛ∫ΛΨ+′ϕdμ+1μΛ∫ΛFz-ϕdμ≤Ψz,or equivalently(19)1μΛ∫ΛΨϕdμ≤1μΛ∫ΛϕΨ+′ϕdμ+Ψz-zμΛ∫ΛΨ+′ϕdμ-1μΛ∫ΛFz-ϕdμ.Taking the infimum over z∈I, we obtain (17).
Proposition 10.
Suppose that all the assumptions of Theorem 8 are satisfied and x¯=1/μ(Λ)∫ΛΨ+′(ϕ)dμ, then(20)0≤Ψ¯-Ψϕ¯-1μΛ∫ΛFϕ-ϕ¯dμ≤infz∈IΨz-zx¯+1μΛ∫ΛϕΨ+′ϕdμ-Ψϕ¯-infz∈I1μΛ∫ΛFz-ϕdμ-1μΛ∫ΛFϕ-ϕ¯dμ≤1μΛ∫ΛϕΨ+′ϕdμ-ϕ¯x¯-infz∈I1μΛ∫ΛFz-ϕdμ-1μΛ∫ΛFϕ-ϕ¯dμ.
Proof.
By setting d1=ϕ¯ and d2=z∈I in (10), we have(21)Ψϕ¯+1μΛ∫ΛFϕ-ϕ¯dμ≤Ψ¯≤Ψz+1μΛ∫Λϕ-zΨ+′ϕdμ-1μΛ∫ΛFz-ϕdμ.Indeed, the following equivalent form of (21) is(22)0≤Ψ¯-Ψϕ¯-1μΛ∫ΛFϕ-ϕ¯dμ≤Ψz+1μΛ∫Λϕ-zΨ+′ϕdμ-Ψϕ¯-1μΛ∫ΛFz-ϕdμ-1μΛ∫ΛFϕ-ϕ¯dμ.Taking the infimum over z∈I, we can easily derive the first and the second inequality in (20). The remaining third inequality in (20) follows because(23)infz∈IΨz-zx¯≤Ψϕ¯-ϕ¯x¯.
Corollary 11.
Let (Λ,X,μ) be a measure space with 0<μ(Λ)<∞ and Ψ:I→R be strongly convex function with modulus c. Suppose ϕ:Λ→I is such that ϕ, Ψ(ϕ), Ψ+′(ϕ), and ϕΨ+′(ϕ)∈L1(μ) and also ϕ¯=1/μ(Λ)∫Λϕdμ, Ψ¯=1/μ(Λ)∫ΛΨ(ϕ)dμ. If d1,d2∈I, then one has(24)Ψd1+Ψ+′d1ϕ¯-d1+cμΛ∫Λϕ-d12dμ≤Ψ¯≤Ψd2+1μΛ∫Λϕ-d2Ψ+′ϕdμ-cμΛ∫Λϕ-d22dμ.
Remark 12.
If we put d1=ϕ¯ in (24) and take probability measure space, then we obtain integral Jensen inequality (9) for strongly convex function.
In the following corollary, we obtain integral Slater’s inequality for strongly convex function.
Corollary 13.
Suppose Ψ, ϕ, μ(Λ), ϕ¯, and Ψ¯ are stated as in Corollary 11 and assume that ∫ΛΨ+′(ϕ)dμ≠0; also if ϕ¯¯=∫ΛϕΨ+′(ϕ)dμ/∫ΛΨ+′(ϕ)dμ∈I, then(25)Ψ¯≤Ψϕ¯¯-cμΛ∫Λϕ-ϕ¯¯2dμ.
Proof.
By setting d2=ϕ¯¯ in (24), we deduced(26)Ψ¯≤Ψ∫ΛϕΨ+′ϕdμ∫ΛΨ+′ϕdμ+1μΛ∫Λϕ-∫ΛϕΨ+′ϕdμ∫ΛΨ+′ϕdμΨ+′ϕdμ-cμΛ∫Λϕ-∫ΛϕΨ+′ϕdμ∫ΛΨ+′ϕdμ2dμ.Since 1/μ(Λ)∫Λϕ-∫ΛϕΨ+′(ϕ)dμ/∫ΛΨ+′(ϕ)dμΨ+′(ϕ)dμ=0, therefore (26) is equivalent to (25).
In the following corollary, we obtain a converse of the Jensen inequality for strongly convex function.
Corollary 14.
Suppose Ψ, ϕ, μ(Λ), ϕ¯, and Ψ¯ are stated as in Corollary 11, then one has(27)0≤Ψ¯-Ψϕ¯-cμΛ∫Λϕ-ϕ¯2dμ≤1μΛ∫Λϕ-ϕ¯Ψ+′ϕdμ-2cμΛ∫Λϕ-ϕ¯2dμ.
Proof.
By setting d1=d2=ϕ¯ in (24), we obtain (27).
3. Jensen-Steffensen Inequality for Riemann-Stieltjes Integrals
To prove the main results of this section, first we prove the following lemma which will play a key role in the proof of main results.
Lemma 15.
Let Ψ:(γ,β)→R be strongly convex function with modulus c. Suppose y is fixed element from interval (γ,β), then
the function Δy(ζ):(γ,β)→R defined by(28)Δyζ=Ψζ-Ψy-Ψ+′yζ-y-cζ-y2
and the function Δ¯y(ζ):(γ,β)→R defined by(29)Δ¯yζ=Ψy-Ψζ-Ψ+′ζy-ζ-cy-ζ2
are nonnegative on (γ,β), decreasing on (γ,y] and increasing on [y,β).
Proof.
By definition of strongly convexity we have (30)Δyζ=Ψζ-Ψy-Ψ+′yζ-y-cζ-y2≥0,∀ζ,y∈γ,βIt means that Δy(ζ) is nonnegative on (γ,β).
Let γ<ζ1<ζ2≤y. Since Ψ is strongly convex with modulus c, therefore Ψ+′(ζ)-2cζ is increasing and hence(31)Ψ+′ζ2-2cζ2≤Ψ+′y-2cy;it follows(32)Ψ+′yζ1-ζ2≤Ψ+′ζ2ζ1-ζ2-2cζ2ζ1-ζ2+2cyζ1-ζ2.Setting ζ=ζ1 and ζ=ζ2 in (28) and then taking the difference, we get(33)Δyζ1-Δyζ2=Ψζ1-Ψζ2-Ψ+′yζ1-ζ2-cζ1-y2+cζ2-y2≥Ψζ1-Ψζ2-Ψ+′ζ2ζ1-ζ2+2cζ2ζ1-ζ2-2cyζ1-ζ2-cζ1-y2+cζ2-y2by using 32=Ψζ1-Ψζ2-Ψ+′ζ2ζ1-ζ2-cζ1-ζ22≥0.Hence, Δy(ζ) is decreasing on (γ,y].
Also, if y≤ζ1<ζ2<β, then similarly as above by strongly convexity we have(34)Ψ+′y-2cy≤Ψ+′ζ1-2cζ1,from which it follows that(35)Ψ+′yζ1-ζ2≥Ψ+′ζ1ζ1-ζ2-2cζ1ζ1-ζ2+2cyζ1-ζ2.Setting ζ=ζ1 and ζ=ζ2 in (28), then taking the difference we get(36)Δyζ1-Δyζ2=Ψζ1-Ψζ2-Ψ+′yζ1-ζ2-cζ1-y2+cζ2-y2≤Ψζ1-Ψζ2-Ψ+′ζ1ζ1-ζ2+2cζ1ζ1-ζ2-2cyζ1-ζ2-cζ1-y2+cζ2-y2by using 35=Ψζ1-Ψζ2-Ψ+′ζ1ζ1-ζ2-cζ1-ζ22≤0.Hence, Δy(ζ) is increasing on [y,β).
In the same manner as above, we also obtain Δ¯y(ζ) is nonnegative on (γ,β).
Let γ<ζ1<ζ2≤y, then by strongly convexity we have(37)Ψ+′ζ1-2cζ1≤Ψ+′ζ2-2cζ2,(38)thatis,Ψ+′ζ2y-ζ2≥Ψ+′ζ1y-ζ2-2cζ1y-ζ2+2cζ2y-ζ2.Setting ζ=ζ1 and ζ=ζ2 in (29) and then taking the difference we get(39)Δ¯yζ1-Δ¯yζ2=Ψζ2-Ψζ1-Ψ+′ζ1y-ζ1+Ψ+′ζ2y-ζ2-cy-ζ12+cy-ζ22≥Ψζ2-Ψζ1-Ψ+′ζ1y-ζ1+Ψ+′ζ1y-ζ2-2cζ1y-ζ2+2cζ2y-ζ2-cy-ζ12+cy-ζ22by 38=Ψζ2-Ψζ1-Ψ+′ζ1ζ2-ζ1-cζ2-ζ12≥0.Hence, Δ¯y(ζ) is decreasing on (γ,y].
Also, if y≤ζ1<ζ2<β, then by strongly convexity we have(40)Ψ+′ζ1-2cζ1≤Ψ+′ζ2-2cζ2,(41)i.e., Ψ+′ζ1y-ζ1≥Ψ+′ζ2y-ζ1-2cζ2y-ζ1+2cζ1y-ζ1.Setting ζ=ζ1 and ζ=ζ2 in (29) and then taking the difference we get(42)Δ¯yζ1-Δ¯yζ2=Ψζ2-Ψζ1-Ψ+′ζ1y-ζ1+Ψ+′ζ2y-ζ2-cy-ζ12+cy-ζ22≤Ψζ2-Ψζ1-Ψ+′ζ2y-ζ1+Ψ+′ζ2y-ζ2-2cζ1y-ζ1+2cζ2y-ζ1-cy-ζ12+cy-ζ22by 41=Ψζ2-Ψζ1-Ψ+′ζ2ζ1-ζ2+cζ1-ζ22≤0.Hence, Δ¯y(ζ) is increasing on [y,β). This completes the proof.
The following lemma is given in [35].
Lemma 16.
Let ϕ:[γ,β]→R be a nonnegative function and suppose ξ:[γ,β]→R is either a bounded variation or continuous. Also assume that the functions ϕ and ξ have no common discontinuity points.
If ϕ is increasing on [γ,β], then(43)ϕβinfγ≤τ≤β∫τβdξx≤∫γβϕxdξx≤ϕβsupγ≤τ≤β∫τβdξx.
If ϕ is decreasing on [γ,β], then(44)ϕγinfγ≤τ≤β∫γτdξx≤∫γβϕxdξx≤ϕγsupγ≤τ≤β∫γτdξx.
In the next result, we prove some general integral inequalities for strongly convex functions.
Theorem 17.
Suppose ϕ:[γ,β]→(κ,ν) is monotonic and continuous function and let Ψ:(κ,ν)→R be a strongly convex function with modulus c. If ξ:[γ,β]→R is either a bounded variation or continuous and satisfying ξ(γ)≤ξ(x)≤ξ(β) for all x∈[γ,β], ξ(β)-ξ(γ)>0, then ϕ¯ and Ψ¯ given by (45)ϕ¯=1ξβ-ξγ∫γβϕxdξx,Ψ¯=1ξβ-ξγ∫γβΨϕxdξxare well defined and ϕ¯∈(κ,ν). Also, if Ψ+′ϕ(x) and ξ have no common discontinuity points, then for d1,d2∈(κ,ν), one has(46)Ψd1+Ψ+′d1ϕ¯-d1+cξβ-ξγ∫γβϕx-d12dξx≤Ψ¯≤Ψd2+1ξβ-ξγ∫γβϕx-d2Ψ+′ϕxdξx-cξβ-ξγ∫γβϕx-d22dξx.
Proof.
Under the given conditions in [35], it has been shown that (47)ϕγ,β=ϕγ,ϕβ⊆κ,ν.We define the function λ:[γ,β]→R by λ(x)=Δd1ϕ(x), where Δd1 is defined as in Lemma 15(a), i.e., (48)λx=Δd1ϕx=Ψϕx-Ψd1-Ψ+′d1ϕx-d1-cϕx-d12.From Lemma 15(a) it follows that λ is nonnegative and since ϕ and Ψ(ϕ) are continuous therefore the integral ∫γβλ(x)dξ(x) exists. Thus we discuss the following three cases:
If ϕ(β)≤d1, since ϕ is increasing on [γ,β] and by Lemma 15(a) Δd1 is decreasing on (κ,d1], therefore λ=Δd1(ϕ(x)) is decreasing on [γ,β]. So using Lemma 16 (b), we obtain (49)∫γβλxdξx≥λγinfγ≤τ≤βξτ-ξγ=0.
If d1≤ϕ(γ), since Δd1 is increasing on [d1,ν) by Lemma 15(a), therefore λ=Δd1(ϕ(x)) is increasing on [γ,β]. So using Lemma 16 (a), we have (50)∫γβλxdξx≥λβinfγ≤τ≤βξβ-ξτ=0.
If ϕ(γ)<d1<ϕ(β), since ϕ is continuous on [γ,β], there exists at least one point x¯∈(γ,β) such that ϕ(x¯)=d1. Also by Lemma 15(a), λ is decreasing on [γ,x¯] and λ is increasing on [x¯,β]. Using Lemma 16, we have (51)∫γβλxdξx=∫γx¯λxdξx+∫x¯βλxdξx≥λγinfγ≤τ≤x¯ξx¯-ξγ+λβinfx¯≤τ≤βξβ-ξx¯=0.
From the above three subcases we conclude that (52)∫γβλxdξx=∫γβΨϕx-Ψd1-Ψ+′d1ϕx-d1-cϕx-d12dξx≥0.(53)i.e., Ψd1ξβ-ξγ+Ψ+′d1∫γβϕxdξx-d1ξβ-ξγ+c∫γβϕx-d12dξx≤∫γβΨϕxdξx.
Dividing (53) by ξ(β)-ξ(γ)>0, we obtain the left side of the inequality (46). Similarly, if ϕ is decreasing we consider the cases ϕ(γ)≤d1 (λ is increasing on [γ,β]), d1≤ϕ(β) (λ is decreasing on [γ,β]), and ϕ(γ)<d1<ϕ(β) (λ is decreasing on [γ,x¯] and increasing on [x¯,β]). In all three cases we obtain ∫γβλ(x)dξ(x)≥0 from the first inequality in (46) which directly follows.
Similarly, we can prove the right side of the inequality (46). We define the function λ¯:[γ,β]→R by λ¯(x)=Δ¯d2(ϕ(x)), where Δ¯d2 is defined as in Lemma 15(b), i.e., (54)λ¯x=Δ¯d2ϕx=Ψd2-Ψϕx-Ψ+′ϕxd2-ϕx-cd2-ϕx2.Since ϕ is monotonic and continuous, Ψ(ϕ) is continuous and Ψ+′(ϕ) is monotonic which have no common discontinuity point with ξ. Therefore the integral ∫γβλ(x)dξ(x) exists. Using the same process as above we have ∫γβλ¯(x)dξ(x)≥0, which means that (55)∫γβΨd2-Ψϕx-Ψ+′ϕxd2-ϕx-cd2-ϕx2dξx≥0.i.e., ξβ-ξγΨd2-∫γβΨϕxdξx+∫γβϕx-d2Ψ+′ϕxdξx-c∫γβϕx-d22dξx≥0.If we divide by ξ(β)-ξ(γ)>0, we obtain the second inequality of (46). So, both inequalities of (46) are proved.
Now, we are in a situation to obtain the following result.
Corollary 18.
Suppose all the assumptions of Theorem 17 are satisfied, then one has(56)0≤Ψ¯-Ψϕ¯-cξβ-ξγ∫γβϕx-ϕ¯2dξx≤1ξβ-ξγ∫γβϕx-ϕ¯Ψ+′ϕxdξx-2cξβ-ξγ∫γβϕx-ϕ¯2dξx.
Proof.
By setting d1=d2=ϕ¯ in (46), we obtain (56).
Remark 19.
If we set c=0 in (56), we obtain Theorem 3.2 in [35].
In the following corollary, we obtain integral Slater’s inequality for strongly convex functions.
Corollary 20.
Suppose all the assumptions of Theorem 17 are satisfied and assume ∫γβΨ+′(ϕ(x))dξ(x)≠0; also if ϕ¯¯=∫γβϕxΨ+′(ϕ(x))dξ(x)/∫γβΨ+′(ϕ(x))dξ(x)∈I, then(57)Ψ¯≤Ψϕ¯¯-cξβ-ξγ∫γβϕx-ϕ¯¯2dξx.
Proof.
Similar to the proof of Corollary 13, setting d2=ϕ¯¯ in the right hand side of (46), we get (57).
Remark 21.
If we set c=0 in (57), we obtain Slater’s inequality for convex functions given in [35].
Data Availability
No data were used to support this study.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
The research was supported by the Natural Science Foundation of China (Grants nos. 61673169 and 11601485).
de FinettiB.Sulle stratificazioni convesse194930417318310.1007/bf02415006MR00434912-s2.0-0000398830PolyakB. T.Existence theorems and convergence of minimizing sequences for extremal problems with constraints196677275HyersD. H.UlamS. M.Approximately convex functions1952382182810.2307/2032186MR0049962Zbl0047.295052-s2.0-8496625633210.1090/S0002-9939-1952-0049962-5PečarićJ. E.ProschanF.TongY. L.1992187Boston, MA, USAAcademic PressJensenJ. L. W. V.Om konvexe funktioner og uligheder mellem middlvaerdier1905164969MangasarianO. L.Pseudo-convex functions19653281290DragomirS. S.Inequalities of Hermite-Hadamard type for φ-convex functions on linear spaces201387http://rgmia.org/papers/v16/v16a87.pdf10.4067/S0716-09172015000400002MR3437222DragomirS. S.Inequalities of Hermite-Hadamard type for λ-convex functions on linear spaces20141718http://rgmia.org/papers/v17/v17a13.pdfVarošanecS.On h-convexity2007326130331110.1016/j.jmaa.2006.02.086MR2277784RajbaT.On strong delta-convexity and Hermite-Hadamard type inequalities for delta-convex functions of higher order201518126729310.7153/mia-18-20MR3277073ChuY.ZhangX.Necessary and sufficient conditions such that extended mean values are Schur-convex or Schur-concave200848122923810.1215/kjm/1250280982MR2437898ElezovićN.PečarićJ.A note on Schur-convex functions200030385385610.1216/rmjm/1021477248MR1797819ShiH.-N.Schur-convex functions related to Hadamard-type inequalities20071112713610.7153/jmi-01-13MR2347711Zbl1131.26017StępniakC.An effective characterization of Schur-convex functions with applications2007141103108MR2310431ZhangX.-M.Schur-convex functions and isoperimetric inequalities1998126246147010.1090/S0002-9939-98-04151-3MR1423343Zbl0908.260092-s2.0-21944445986AnastassiouG. A.Basic and s-convexity Ostrowski and Grüss type inequalities involving several functions2013172189212MR3135080SetE.ÖzdemirM. E.DragomirS. S.On Hadamard-type inequalities involving several kinds of convexity201020101228684510.1155/2010/286845MR2718494YangZ.-H.QianW.-M.ChuY.-M.ZhangW.On approximating the error function201821246947910.7153/mia-2018-21-32MR3776087YangZ.-H.QianW.-M.ChuY.-M.ZhangW.On approximating the arithmetic-geometric mean and complete elliptic integral of the first kind201846221714172610.1016/j.jmaa.2018.03.005MR3774313Zbl06864634RobertsA. W.VarbergD. E.1973London, UKAcademic PressMR0442824Zbl0271.26009AdamekM.On a problem connected with strongly convex functions20161941287129310.7153/mia-19-94MR3571583MoradiH. R.OmidvarM. E.Adil KhanM.NikodemK.Around Jensen's inequality for strongly convex functions2018921253710.1007/s00010-017-0496-5MR3745678SlaterM. L.A companion inequality to Jensen's inequality198132216016610.1016/0021-9045(81)90112-XMR633700Zbl0521.41011Adil KhanM.PečarićJ.On Slater's integral inequality20115223124110.7153/jmi-05-21MR2817207Adil KhanM.PečarićJ. E.Improvement and reversion of Slater's inequality and related results201020101464603410.1155/2010/646034MR2665487DragomirS. S.GohC. J.A counterpart of Jensen's discrete inequality for differentiable convex mappings and applications in information theory199624211110.1016/0895-7177(96)00085-4MR1403525Zbl0862.94010MatićM.PečarićJ.Some companion inequalities to Jensen's inequality20003335536810.7153/mia-03-36MR1768496MerentesN.NikodemK.Remarks on strongly convex functions2010801-219319910.1007/s00010-010-0043-0MR2736950Zbl1214.260072-s2.0-78149495610Klaričić BakulaM.NikodemK.On the converse Jensen inequality for strongly convex functions2016434151652210.1016/j.jmaa.2015.09.032MR3404570Adil KhanM.KhanJ.PečarićGeneralization of Jensen's and Jensen-Steffensen's inequalities by generalized majorization theorem20171141049107410.7153/jmi-2017-11-80MR3711393DragomirS. S.Adil KhanM.AbathunA.Refinement of the Jensen integral inequality20161422122810.1515/math-2016-0020MR3491891Zbl1347.26047Mitroi-SymeonidisF.-C.MinculeteN.On the Jensen functional and strong convexity201841131131910.1007/s40840-015-0293-zMR3743840NikodemK.2014New York, NY, USASpringerNikodemK.PálesZ.Characterizations of inner product spaces by strongly convex functions201151838710.15352/bjma/1313362982MR2738522Klaričić BakulaM.MatićM.PečarićJ.Generalizations of the Jensen-Steffensen and related inequalities20097478780310.2478/s11533-009-0052-1MR2563450