{
  "id": "concave-function",
  "title": "Concave function",
  "lang": "en",
  "source_lang": "en",
  "summary": "A concave function takes values at weighted averages that are at least as large as the corresponding weighted averages of its values.",
  "category": "mathematics",
  "subcategory": "analysis",
  "type": "concept",
  "aliases": [
    "Concave-down function",
    "Concave downward function"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/concave-function/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/concave-function/",
      "json": "https://aiwiki.page/zh/concave-function.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A concave function is a real-valued function whose value at any weighted average of two points is at least the same weighted average of its values at those points. Geometrically, its graph lies on or above every line segment joining two points on the graph. Concavity is the sign-reversed counterpart of convexity: \\(f\\) is concave exactly when \\(-f\\) is a convex function. It is a central concept in optimization, probability, economics, and mathematical analysis."
    },
    {
      "heading": "Definition",
      "text": "Let \\(C\\) be a convex subset of a real vector space. A function \\(f:C\\to\\mathbb{R}\\) is concave if, for every \\(x,y\\in C\\) and every \\(t\\in[0,1]\\),\n\n\\[f((1-t)x+ty)\\geq (1-t)f(x)+tf(y).\\]\n\nThe requirement that \\(C\\) be convex ensures that \\((1-t)x+ty\\) belongs to the domain.\n\nMore generally, the two-point inequality implies that for any finite collection \\(x_1,\\ldots,x_m\\in C\\) and nonnegative weights \\(\\lambda_1,\\ldots,\\lambda_m\\) satisfying \\(\\sum_i\\lambda_i=1\\),\n\n\\[f\\left(\\sum_{i=1}^{m}\\lambda_i x_i\\right) \\geq \\sum_{i=1}^{m}\\lambda_i f(x_i).\\]\n\nThe argument on the left is a convex combination of the points. Thus concavity means that taking an average before applying the function produces a value no smaller than averaging afterward.\n\nConcavity concerns the shape of a function, not whether it increases or decreases. A concave function may be increasing, decreasing, or increasing on one part of its domain and decreasing on another.\n\nStrict concavity\n\nThe function is strictly concave if\n\n\\[f((1-t)x+ty)>(1-t)f(x)+tf(y)\\]\n\nwhenever \\(x\\ne y\\) and \\(0<t<1\\).\n\nStrict concavity rules out every nontrivial straight segment in the graph. An affine function, such as \\(f(x)=a^\\mathsf{T}x+b\\), is both concave and convex, but is not strictly concave on any domain containing two distinct points.\n\nStrong concavity\n\nOn a convex subset of a Euclidean space, \\(f\\) is strongly concave with parameter \\(\\mu>0\\) if\n\n\\[f((1-t)x+ty) \\geq (1-t)f(x)+tf(y) +\\frac{\\mu}{2}t(1-t)\\|x-y\\|^2\\]\n\nfor all admissible \\(x,y,t\\), where \\(\\|\\cdot\\|\\) is the Euclidean norm.\n\nEquivalently,\n\n\\[x\\longmapsto f(x)+\\frac{\\mu}{2}\\|x\\|^2\\]\n\nis concave. Strong concavity supplies a uniform quadratic lower bound on the gap between the graph and its chords. It implies strict concavity, but strict concavity need not imply strong concavity."
    },
    {
      "heading": "Geometric characterizations",
      "text": "Chords and hypographs\n\nFor a function of one variable, concavity means that each chord of the graph lies on or below the graph between its endpoints.\n\nIn any dimension, the hypograph of \\(f\\) is\n\n\\[\\operatorname{hypo} f = \\{(x,r):x\\in C,\\ r\\leq f(x)\\}.\\]\n\nA function is concave if and only if its hypograph is a convex set. This characterization expresses concavity entirely in terms of set geometry and does not require differentiability.\n\nTangent upper bounds\n\nSuppose \\(f\\) is differentiable on an open convex subset of \\(\\mathbb{R}^n\\). Then \\(f\\) is concave if and only if\n\n\\[f(y)\\leq f(x)+\\nabla f(x)^\\mathsf{T}(y-x)\\]\n\nfor every \\(x,y\\) in its domain, where \\(\\nabla f(x)\\) is its gradient.\n\nThus the tangent affine function at any point is a global upper bound. This is the reverse of the tangent lower-bound property of convex functions.\n\nFor a function that is not differentiable, a vector \\(g\\) satisfying\n\n\\[f(y)\\leq f(x)+g^\\mathsf{T}(y-x) \\qquad\\text{for all }y\\in C\\]\n\nis called a supergradient at \\(x\\). For a finite concave function on a finite-dimensional convex domain, a supporting affine upper bound exists at every point in the relative interior of the domain; existence at boundary points requires additional care."
    },
    {
      "heading": "Derivative criteria",
      "text": "Functions of one variable\n\nA differentiable function on an open interval is concave if and only if its derivative is nonincreasing. If it is twice differentiable, this is equivalent to\n\n\\[f''(x)\\leq 0\\]\n\nthroughout the interval.\n\nThe stronger condition\n\n\\[f''(x)<0\\]\n\nthroughout the interval guarantees strict concavity, but is not necessary for it. For example, \\(f(x)=-x^4\\) is strictly concave on \\(\\mathbb{R}\\), although \\(f''(0)=0\\).\n\nConcavity can also be described without derivatives. For \\(a<b<c\\) in the domain,\n\n\\[\\frac{f(b)-f(a)}{b-a} \\geq \\frac{f(c)-f(a)}{c-a} \\geq \\frac{f(c)-f(b)}{c-b}.\\]\n\nThese inequalities express the decrease of secant slopes as the interval moves to the right.\n\nFunctions of several variables\n\nFor a twice continuously differentiable function on an open convex subset of \\(\\mathbb{R}^n\\), concavity is equivalent to the Hessian matrix being negative semidefinite everywhere:\n\n\\[v^\\mathsf{T}\\nabla^2f(x)v\\leq 0 \\qquad\\text{for every }v\\in\\mathbb{R}^n.\\]\n\nEquivalently, every eigenvalue of the Hessian is nonpositive.\n\nA negative definite Hessian everywhere guarantees strict concavity. As in one dimension, this condition is sufficient but not necessary.\n\nUnder the same smoothness assumptions, strong concavity with parameter \\(\\mu\\) is equivalent to\n\n\\[v^\\mathsf{T}\\nabla^2f(x)v \\leq -\\mu\\|v\\|^2\\]\n\nfor every \\(x\\) and \\(v\\). If the domain lies in a lower-dimensional affine subspace, the relevant curvature conditions apply only along directions parallel to that subspace."
    },
    {
      "heading": "Examples",
      "text": "- Logarithm. The natural logarithm is strictly concave on \\((0,\\infty)\\), since\n\n\\[\\frac{d^2}{dx^2}\\log x=-\\frac{1}{x^2}<0.\\]\n\nIt is not strongly concave on the entire positive half-line because this curvature approaches zero as \\(x\\to\\infty\\).\n- Fractional powers. For \\(0<\\alpha<1\\), the function \\(f(x)=x^\\alpha\\) is strictly concave on \\([0,\\infty)\\). The square-root function is a familiar example.\n- Negative quadratic functions. If \\(Q\\) is symmetric and positive semidefinite, then\n\n\\[f(x)=-\\frac12x^\\mathsf{T}Qx+b^\\mathsf{T}x+c\\]\n\nis concave. If \\(Q\\) is positive definite, the function is strongly concave, with parameter equal to the smallest eigenvalue of \\(Q\\).\n- Negative absolute value. The function \\(f(x)=-|x|\\) is concave on \\(\\mathbb{R}\\), but is neither differentiable at zero nor strictly concave on the whole domain.\n- Minimum of coordinates. The function\n\n\\[f(x_1,\\ldots,x_n)=\\min\\{x_1,\\ldots,x_n\\}\\]\n\nis concave and generally nondifferentiable where several coordinates attain the minimum.\n- Entropy. For a finite probability vector \\(p=(p_1,\\ldots,p_n)\\), Shannon entropy\n\n\\[H(p)=-\\sum_{i=1}^{n}p_i\\log p_i,\\]\n\nwith \\(0\\log 0=0\\), is strictly concave on the probability simplex."
    },
    {
      "heading": "Preservation and regularity properties",
      "text": "Several operations preserve concavity:\n\n- A sum of concave functions with nonnegative coefficients is concave.\n- Adding an affine function preserves concavity.\n- Composing a concave function with an affine map preserves concavity on the resulting convex domain.\n- If \\(g\\) is concave and nondecreasing, and \\(f\\) is concave, then \\(g\\circ f\\) is concave wherever the composition is defined.\n- The pointwise infimum of a family of concave functions on a common convex domain is concave, provided it is finite-valued there.\n\nBy contrast, the pointwise maximum of concave functions need not be concave. For example, \\(x\\) and \\(-x\\) are concave, but their maximum \\(|x|\\) is not. Products of concave functions also do not generally preserve concavity.\n\nA finite concave function on a finite-dimensional convex domain is continuous on the relative interior of its domain and is locally Lipschitz there. Continuity at boundary points does not follow from concavity alone. For example, the function on \\([0,1]\\) defined by\n\n\\[f(0)=0,\\qquad f(x)=1\\quad\\text{for }0<x\\leq1\\]\n\nis concave but discontinuous at zero."
    },
    {
      "heading": "Jensen’s inequality",
      "text": "Jensen’s inequality extends the finite weighted-average inequality to probability distributions. If \\(X\\) is an integrable random vector taking values in a convex domain, and the relevant expectations are finite, then\n\n\\[f(\\mathbb{E}[X])\\geq\\mathbb{E}[f(X)].\\]\n\nHere \\(\\mathbb{E}\\) denotes expectation. The inequality compares the function evaluated at the mean with the mean of the function’s values.\n\nIf \\(f\\) is strictly concave, equality holds only when \\(X\\) is constant almost surely. For a general concave function, equality may also occur when the distribution is supported on a region where the function is affine.\n\nFor the logarithm, the finite form gives\n\n\\[\\log\\left(\\sum_i\\lambda_i x_i\\right) \\geq \\sum_i\\lambda_i\\log x_i,\\]\n\nwhich, after exponentiation, yields the weighted arithmetic–geometric mean inequality."
    },
    {
      "heading": "Role in optimization",
      "text": "Maximizing a concave objective function over a convex feasible set is a standard form of convex optimization, because it is equivalent to minimizing the convex function \\(-f\\).\n\nConcavity gives several important guarantees:\n\n- Every local maximum is a global maximum.\n- The set of maximizers is convex.\n- A strictly concave function has at most one maximizer on a convex domain.\n- For a differentiable concave function, an interior point with zero gradient is a global maximizer.\n\nMore generally, a feasible point \\(x^\\ast\\) is a global maximizer if\n\n\\[\\nabla f(x^\\ast)^\\mathsf{T}(y-x^\\ast)\\leq0 \\qquad\\text{for all feasible }y.\\]\n\nThese statements do not guarantee that a maximizer exists. For example, \\(\\log x\\) is concave on \\((0,\\infty)\\) but unbounded above. A continuous function on a nonempty compact feasible set attains a maximum by the extreme value theorem.\n\nStrong concavity additionally gives a quantitative bound. At an unconstrained maximizer \\(x^\\ast\\) of a differentiable, \\(\\mu\\)-strongly concave function,\n\n\\[f(x^\\ast)-f(x)\\geq\\frac{\\mu}{2}\\|x-x^\\ast\\|^2.\\]\n\nThus a small objective-value gap bounds the distance to the optimizer."
    },
    {
      "heading": "Applications and related concepts",
      "text": "In economics, concave utility functions model diminishing marginal utility: where derivatives exist, marginal utility does not increase as consumption rises. In expected-utility models, concavity also produces the inequality underlying risk aversion through Jensen’s inequality.\n\nIn maximum likelihood estimation, some statistical models have concave log-likelihood functions. When this property holds on a convex parameter domain, it simplifies optimization and rules out nonglobal local maxima. Concavity of a log-likelihood is a model-dependent property, not a general property of likelihoods.\n\nIn information theory, entropy concavity describes the effect of mixing probability distributions: the entropy of a mixture is at least the corresponding average of the component entropies.\n\nA related but weaker property is quasiconcavity. A function is quasiconcave if\n\n\\[f((1-t)x+ty)\\geq\\min\\{f(x),f(y)\\}.\\]\n\nEquivalently, all of its upper level sets are convex. Every concave function is quasiconcave, but the converse fails. For example, the exponential function on \\(\\mathbb{R}\\) is quasiconcave because it is increasing, yet it is strictly convex rather than concave. Concavity imposes quantitative restrictions on function values along segments, whereas quasiconcavity constrains only the geometry of upper level sets."
    }
  ],
  "links": [
    "mathematical-function",
    "convex-function",
    "convex-set",
    "vector-space",
    "convex-combination",
    "affine-map",
    "euclidean-space",
    "gradient",
    "derivative",
    "hessian-matrix",
    "natural-logarithm",
    "information-entropy",
    "continuous-function",
    "jensens-inequality",
    "random-variable",
    "expected-value",
    "objective-function",
    "feasible-set",
    "convex-optimization",
    "extreme-value-theorem",
    "economics",
    "maximum-likelihood-estimation"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}