Fact-checked by Grok 2 weeks ago

Set-valued function

A set-valued function, also known as a multifunction or correspondence, is a mathematical mapping F: X \to 2^Y from a domain set X to the power set of a codomain set Y, where for each x \in X, F(x) is typically a nonempty subset of Y, allowing multiple possible outputs for a single input rather than a single value as in ordinary functions.^[1] This generalization extends classical function theory to scenarios where outcomes are inherently uncertain or multiple, such as in modeling choices or relations in complex systems.^[2] Set-valued functions play a central role in various branches of mathematics, including topology, analysis, and optimization, where they facilitate the study of properties like continuity (upper and lower hemicontinuity) and fixed points. A key result is Kakutani's fixed-point theorem, which guarantees the existence of fixed points for upper hemicontinuous set-valued functions on compact, convex subsets of Euclidean spaces, generalizing Brouwer's theorem for single-valued functions and enabling proofs of equilibrium existence in many models.^[3] In applications, set-valued functions are essential in economics for representing preferences, budget correspondences, and general equilibrium theory, as well as in control theory for differential inclusions and in variational analysis for optimization problems with set constraints.^[4] They also underpin integrals like the Aumann integral, which integrates selections from set-valued mappings to handle uncertainty in stochastic processes and economic modeling.^[5]

Definition and Fundamentals

Formal Definition

A set-valued function, also referred to as a multifunction or correspondence, is a mathematical mapping that assigns to each element in a domain set a collection of elements from a codomain set, rather than a single element. Formally, given two sets X and Y, a set-valued function F: X \to 2^Y is defined such that for every x \in X, F(x) is a subset of Y, where $2^Y denotes the power set of Y, the collection of all subsets of Y. This allows F(x) to be empty, a singleton, or any finite or infinite subset of Y, generalizing the concept of ordinary functions. The domain of F is typically taken as the set \{x \in X \mid F(x) \neq \emptyset\}, which excludes points where the function yields the empty set, ensuring that the mapping is non-vacuous at each domain element. The range, or image, of F is then the union \bigcup_{x \in X} F(x), representing all elements in Y that are attained by at least one x. Cases where F(x) = \emptyset for some x imply that such x may be omitted from the effective domain, which can affect applications in optimization or analysis by restricting the feasible set. Single-valued functions serve as a special case of set-valued functions, where |F(x)| = 1 for all x in the domain, reducing the power set mapping to a standard function f: X \to Y. This inclusion highlights how set-valued functions extend classical function theory to handle ambiguities or multiple outcomes, such as in differential inclusions or economic modeling.

Notation and Terminology

Set-valued functions are commonly denoted using a special arrow to indicate that the image of each point in the domain is a set rather than a single element. The notation F: X \rightrightarrows Y signifies a set-valued map from a set X to subsets of a set Y, where for each x \in X, F(x) \subseteq Y. This double-arrow convention distinguishes set-valued maps from single-valued functions, which use the standard F: X \to Y. The graph of a set-valued map F: X \rightrightarrows Y, denoted \mathrm{Gr} F, is defined as the set \mathrm{Gr} F = \{(x,y) \in X \times Y \mid y \in F(x)\}. This graphical representation embeds the set-valued map into a subset of the product space X \times Y, facilitating the study of its topological and analytical properties. Several synonymous terms are used interchangeably for set-valued functions in the literature, including "multifunction," "point-to-set map," and "correspondence." The term "multifunction" emphasizes the multiple values assigned to each input. "Point-to-set map" highlights the mapping from points in the domain to sets in the codomain. The term "correspondence" derives from mathematical economics, where it describes relations such as demand or budget sets in general equilibrium theory, as introduced in foundational works on economic equilibrium.^[6] Conventions for specifying properties of the images F(x) are standard in set-valued analysis. A set-valued map F: X \rightrightarrows Y is closed-valued if F(x) is a closed subset of Y for every x \in X. Similarly, F is convex-valued if F(x) is convex for each x, and compact-valued if F(x) is compact, where compactness assumes a suitable topology on Y. These properties ensure that the images possess desirable geometric or topological features essential for theorems in optimization and viability theory. The inverse of a set-valued map F: X \rightrightarrows Y is defined as the set-valued map [F^{-1}](/page/Inverse): Y \rightrightarrows X given by F^{-1}(y) = \{x \in X \mid y \in F(x)\}. This construction reverses the direction of the mapping while preserving the set-valued nature, and it plays a key role in duality and fixed-point arguments.^[7]

Distinctions and Relations

Relation to Multivalued Functions

The terms "multivalued function" and "set-valued function" are synonymous in contemporary mathematics, both referring to mappings from a domain to subsets of a codomain, though the former historically emphasized the assignment of multiple discrete values to inputs in specific analytical contexts.^[8]^[9] The term "multivalued function" originated in 19th-century complex analysis, where it described functions like the square root that yield multiple values due to branching in the complex plane, prompting Bernhard Riemann's 1851 introduction of Riemann surfaces to resolve such ambiguities by constructing multi-sheeted coverings of the domain.^[10] This usage persisted into the early 20th century, as seen in definitions equating multivalued functions with correspondences yielding sets of outputs for each input.^[11] In contrast, "set-valued function" gained prominence post-1950s within set theory, functional analysis, and optimization, reflecting a more general framework where mappings explicitly target the power set of the codomain to accommodate non-smooth phenomena like subdifferentials.^[8]^[4] The coexistence of terms stems from disciplinary preferences: "multivalued" retains historical ties to complex analysis and algebraic multiplicity, while "set-valued" (or "multifunction") avoids conflation with purely algebraic interpretations and aligns with variational methods in optimization, as developed by figures like Jean-Jacques Moreau in the 1960s.^[12] There is no substantive mathematical distinction between the two; both denote relations where the image of each element is a subset, formally F(x) ⊆ Y for domain X and codomain Y.^[8]

Comparison to Single-Valued Functions

A single-valued function f: X \to Y constitutes a special case of a set-valued function F: X \to 2^Y, wherein F(x) = \{ f(x) \} for all x \in X, ensuring the image is always a singleton set. This inclusion allows classical results on single-valued functions to be reinterpreted within the more general set-valued framework, facilitating the extension of analytical tools to broader contexts.^[13] Key differences arise from the nature of the images in set-valued functions, which may be empty, contain multiple elements, or be infinite, unlike the unique output of single-valued functions. This variability impacts fundamental operations: invertibility, for example, is more nuanced, as set-valued functions behave like relations and typically lack a unique single-valued inverse even if they are "bijective" in a set-theoretic sense. Composition also requires adjustment; the standard definition (F \circ G)(x) = \bigcup_{y \in G(x)} F(y) involves a union over the intermediate image set G(x), contrasting with the direct point evaluation f(g(x)) for single-valued functions and potentially yielding larger or more complex output sets.^[13]^[14] When evaluating set-valued functions, there is no unique "value" at each input x; instead, one determines membership y \in F(x), shifting analysis from equality to inclusion relations. This perspective underscores their role as a generalization of single-valued functions, particularly in modeling uncertainty—where the output set represents possible outcomes—or multiple choices, as in economic theory where correspondences depict feasible or optimal sets under varying constraints.^[13]^[15]^[16]

Key Properties

Graph and Domain Concepts

The graph of a set-valued function F: X \rightrightarrows Y, where X and Y are sets, is defined as the subset \mathrm{Gr}\, F \subseteq X \times Y given by

\mathrm{Gr}\, F = \{ (x, y) \in X \times Y \mid y \in F(x) \}.

This construction embeds the set-valued mapping into the product space, allowing the use of topological or analytical tools on the graph itself.^[13] The projection of the graph onto the domain space, \pi_X(\mathrm{Gr}\, F) = \{ x \in X \mid F(x) \neq \emptyset \}, defines the effective domain (or natural domain) of F, which consists of all points where F is nonempty.^[13] A set-valued function is called total if its effective domain coincides with the entire space X, meaning F(x) \neq \emptyset for all x \in X. In topological spaces, key properties of the graph provide foundational insights into the behavior of set-valued functions. Specifically, F is said to have a closed graph if \mathrm{Gr}\, F is a closed subset of the product topology on X \times Y.^[13] This closedness condition implies that F is closed-valued, meaning F(x) is a closed subset of Y for every x in the effective domain, as limits of sequences in the graph project to limits in the values.^[13] Conversely, closed-valued functions do not necessarily have closed graphs, as the latter requires uniformity across the domain.^[13] For set-valued functions taking values in ordered or lattice-structured spaces (such as partially ordered sets or power sets under inclusion), analogs of epigraphs and hypographs extend scalar concepts to capture "upper" and "lower" sets relative to F. In such settings, the epigraph of F is defined as

\mathrm{epi}\, F = \{ (x, A) \in X \times 2^Y \mid A \supseteq F(x) \},

comprising pairs where A is a superset of the image F(x), reflecting points "above" F in the lattice of subsets.^[13] Dually, the hypograph is

\mathrm{hyp}\, F = \{ (x, A) \in X \times 2^Y \mid A \subseteq F(x) \},

capturing subsets contained within F(x). These constructions are particularly useful in ordered environments for studying convexity or viability properties of set-valued mappings.^[13]

Continuity Notions

In the context of set-valued functions, continuity is generalized from single-valued functions to account for the multiplicity of outputs, requiring notions that capture the behavior of entire images under limits. These concepts, primarily upper and lower semicontinuity (also termed upper and lower hemicontinuity), were formalized in the mid-20th century to extend topological properties to multifunctions.^[17] Upper semicontinuity at a point x_0 in the domain means that the limit superior of the images F(x) as x approaches x_0 is contained in F(x_0), i.e., \limsup_{x \to x_0} F(x) \subseteq F(x_0).^[13] Equivalently, for compact-valued maps in metric spaces, this holds if the graph of F is closed at (x_0, y) for every y \in F(x_0), or if \sup_{y \in F(x)} \inf_{z \in F(x_0)} d(y, z) \to 0 as x \to x_0; this is the upper part of the Hausdorff distance, aligning upper hemicontinuity with the corresponding semi-metric for closed sets.^[13] Lower semicontinuity at x_0 ensures that sequences in the images can approximate points in F(x_0); specifically, for every y_0 \in F(x_0) and sequence x_n \to x_0, there exist y_n \in F(x_n) such that y_n \to y_0.^[17] This is equivalent to the graph of F being open at (x_0, y_0) relative to its projection, or, in terms of the Hausdorff metric for compact sets, \sup_{z \in F(x_0)} \inf_{y \in F(x)} d(z, y) \to 0 as x \to x_0; this is the lower part of the Hausdorff distance.^[13] The Hausdorff distance itself, defined for nonempty subsets A, B of a metric space as

d_H(A, B) = \max\left\{ \sup_{a \in A} \inf_{b \in B} d(a, b), \sup_{b \in B} \inf_{a \in A} d(a, b) \right\},

provides a metric on the space of closed bounded sets, enabling convergence analysis for set-valued maps.^[13] A set-valued function is fully continuous at x_0 if it is both upper and lower semicontinuous, implying continuity with respect to the Hausdorff metric when values are compact and nonempty.^[13] This full continuity facilitates applications such as fixed-point theorems; for instance, Kakutani's theorem guarantees a fixed point for an upper semicontinuous, convex-valued map on a compact convex set in Euclidean space.^[18]

Examples and Illustrations

Basic Examples

A fundamental illustration of a set-valued function arises from generalizing the absolute value function to account for both positive and negative branches. Consider the mapping F: \mathbb{R} \to 2^{\mathbb{R}} defined by F(x) = \{ -|x|, |x| \} for all x \in \mathbb{R}. At x = 0, F(0) = \{ 0 \}, yielding a single value, but for any x \neq 0, F(x) consists of two distinct points symmetric about zero, demonstrating how set-valued functions can assign multiple outputs to inputs outside specific points. This example highlights the multivalued nature inherent in the formal definition of set-valued mappings, where the image of each element is a non-empty subset of the codomain. Another basic example involves interval-valued outputs, which produce connected sets as images. Define F: [0, \infty) \to 2^{\mathbb{R}} by F(x) = [x, x+1] for x \geq 0. Here, for each input x, the output is the closed interval from x to x+1, forming a continuum of possible values rather than discrete points. This construction illustrates how set-valued functions can model uncertainty or ranges in a structured manner, with the image sets being compact and convex subsets of \mathbb{R}. Such interval mappings are commonly used to build intuition for more complex set-valued structures. Constant set-valued functions provide a simple case where the output does not depend on the input, emphasizing independence from the domain elements. For instance, let F: \mathbb{R} \to 2^{\mathbb{R}} be given by F(x) = \{ 0, 1 \} for all x \in \mathbb{R}. Regardless of the value of x, the image remains the fixed two-point set \{ 0, 1 \}, which reveals the non-injective character of such functions since distinct inputs map to identical outputs. This example underscores the potential for set-valued mappings to represent invariant relations across the entire domain. Set-valued functions may also incorporate empty images for certain inputs, defining partial domains where the mapping is undefined. Consider F: \mathbb{R} \to 2^{\mathbb{R}} such that F(x) = \emptyset for x < 0 and F(x) = \{ x \} for x \geq 0. The effective domain is then [0, \infty), as the empty set indicates no output for negative inputs, transitioning to single-valued behavior on the non-negative reals. This partial definition illustrates how set-valued functions can extend single-valued ones while allowing for undefined regions, aligning with the domain concept in the formal framework.

Geometric Examples

One geometric construction of a set-valued function arises from the projection onto a closed convex set C \subset \mathbb{R}^n, defined as

F(x) = \left\{ y \in C \;\middle|\; \|x - y\| = \inf_{z \in C} \|x - z\| \right\}.

This set consists of all points in C at the minimal distance from x, known as the nearest-point projection. For a closed convex C in Euclidean space, F(x) is a singleton due to the uniqueness of the projection, but the formulation highlights its set-valued nature in broader contexts, such as non-Hilbert spaces or generalized distances where multiple closest points may exist. Geometrically, F(x) corresponds to the intersection of C with the sphere centered at x of radius \mathrm{dist}(x, C), providing a visual representation of how the function "selects" boundary points of C relative to x.^[19] Another illustrative example involves upper and lower envelopes, where a set-valued function models uncertainty by mapping to intervals between two single-valued functions. Specifically, let f, g: \mathbb{R} \to \mathbb{R} satisfy f(x) \leq g(x) for all x in the domain; then F(x) = [f(x), g(x)], an interval-valued function that is a special case of set-valued functions. Geometrically, in the plane, the graph of F forms a filled region bounded below by the curve y = f(x) and above by y = g(x), resembling a tubular band that captures a range of possible outputs at each x, useful for visualizing bounded uncertainty in parameters or measurements.^[20] The Minkowski sum offers a spatial construction for combining set-valued functions, defined as F(x) = G(x) + H(x) = \{ g + h \mid g \in G(x),\ h \in H(x) \} for set-valued maps G, H: \mathbb{R}^n \rightrightarrows \mathbb{R}^m. This operation translates every point in one set by every point in the other, yielding a new set that expands the geometric extent. For instance, if G(x) and H(x) are line segments in \mathbb{R}^2, F(x) forms a zonotope such as a parallelogram, illustrating how the sum "sweeps" one segment along the other to fill a convex polygon; if the input sets are disks, the sum is a larger disk, emphasizing additive growth in area and shape.^[19] A further geometric example demonstrates branching structures through directed segments in \mathbb{R}^2, where F(t) = \{ s (\cos t, \sin t) \mid 0 \leq s \leq 1 \} for t \in \mathbb{R}, mapping each angle t to the unit line segment emanating from the origin in that direction. The graph of F over an interval of t traces a conical surface in \mathbb{R}^3 (with coordinates (t, s \cos t, s \sin t)), while the image F(\mathbb{R}) fills the unit disk; however, selections from F can produce branching paths, such as multiple rays diverging from the origin, highlighting how set-valued assignments generate non-convex images or filled regions in applications like reachable sets in dynamics.^[19]

Theoretical Framework

Set-Valued Analysis

Set-valued analysis is a branch of mathematics that extends classical analysis to functions taking values in sets rather than single points, providing tools to handle uncertainties, multivalued mappings, and non-smooth phenomena in various fields. It encompasses concepts like measurability, convergence of sets, integration, and differential equations adapted to set-valued settings, building on topological and measure-theoretic foundations to study properties such as limits and derivatives of multifunctions. This framework is essential for generalizing deterministic models to include ambiguities or choices, where outcomes are sets rather than unique values. A fundamental notion in set-valued analysis is the measurability of a set-valued function F: X \to 2^Y, where X and Y are measurable spaces. F is measurable if its graph \operatorname{Gr} F = \{(x, y) \in X \times Y \mid y \in F(x)\} is measurable in the product space X \times Y. This graph measurability ensures that projections and selections behave compatibly with the underlying measure, allowing integration and limits to be well-defined; it presupposes weaker forms of continuity, such as upper or lower hemicontinuity, but extends beyond them to handle non-continuous cases. Convergence modes for sets are crucial in set-valued analysis to extend pointwise limits to multifunctions. The Painlevé-Kuratowski convergence, also known as Kuratowski convergence, defines the limit superior and inferior of a sequence of sets F_n converging to F such that \limsup F_n = \bigcap_{n} \overline{\bigcup_{k \geq n} F_k} and \liminf F_n = \bigcup_{n} \overline{\bigcap_{k \geq n} F_k}, with convergence occurring when these coincide with F. This topology on the space of closed sets captures both set inclusion and Hausdorff-like proximity, enabling the study of sequential limits for graphs and epigraphs in optimization and dynamics. Integration of set-valued functions is formalized through the Aumann integral, which for a measurable F: (T, \mathcal{B}, \mu) \to 2(\mathbb{R}^m) is defined as \int F \, d\mu = \left\{ \int f \, d\mu \;\middle|\; f \text{ is an integrable measurable selection of } F \right\}, where a measurable selection f satisfies f(t) \in F(t) for \mu-almost all t \in T.^[21] This integral generalizes the Lebesgue integral by integrating over selectable functions, ensuring the result is a compact convex set when F has closed convex values and is integrably bounded, and it coincides with the classical integral when F is single-valued.^[21] Differential inclusions extend ordinary differential equations to set-valued settings, replacing \dot{x} = f(t, x) with \dot{x}(t) \in F(t, x(t)), where F is a set-valued map. Solutions are absolutely continuous functions x: [0, T] \to \mathbb{R}^n satisfying the inclusion almost everywhere, generalizing trajectories to tubes of possible evolutions under uncertainty or control. Existence and regularity of solutions rely on properties like measurability and upper semicontinuity of F, with the solution set forming a viable set in the state space.

Selection and Fixed-Point Theorems

Selection theorems address the existence of single-valued continuous functions that pick elements from the images of set-valued functions, while fixed-point theorems guarantee points fixed by the set-valued map itself. These results are foundational in set-valued analysis, providing existence guarantees under topological and convexity assumptions. Michael's selection theorem establishes conditions for the existence of continuous selections for set-valued maps into Banach spaces. Specifically, if F: X \to 2^Y is a lower hemicontinuous map from a paracompact topological space X to a Banach space Y, with nonempty, closed, and convex values F(x) \subseteq Y for each x \in X, then there exists a continuous selection f: X \to Y such that f(x) \in F(x) for all x \in X.^[22] Metric spaces, being paracompact, satisfy the domain condition, making the theorem applicable in many practical settings involving complete normed spaces. Kakutani's fixed-point theorem extends Brouwer's fixed-point theorem to set-valued maps, ensuring the existence of fixed points under suitable continuity and convexity conditions. For an upper hemicontinuous set-valued map F: K \rightrightarrows K, where K is a nonempty, compact, convex subset of Euclidean space, with nonempty, convex, and compact values F(x) \subseteq K for each x \in K, there exists some x \in K such that x \in F(x).^[18] Upper hemicontinuity here means that for every open set V containing F(x), there is a neighborhood U of x such that F(u) \subseteq V for all u \in U. Brouwer's fixed-point theorem serves as a special case of Kakutani's result for single-valued continuous functions. It states that any continuous function f: K \to K, where K is a nonempty, compact, convex set in \mathbb{R}^n, has at least one fixed point x \in K with f(x) = x. This single-valued version highlights how set-valued theorems generalize classical fixed-point results. Berge's maximum theorem provides continuity properties for optimal value functions and their argmax sets in parametric optimization problems involving set-valued constraints. If u: S \times T \to \mathbb{R} is continuous, and \Gamma: S \rightrightarrows T is a nonempty, compact-valued, upper hemicontinuous correspondence between topological spaces, then the value function v(s) = \max_{t \in \Gamma(s)} u(s, t) is continuous on S, and the argmax correspondence \Phi(s) = \arg\max_{t \in \Gamma(s)} u(s, t) is nonempty, compact-valued, and upper hemicontinuous. This theorem ensures stability of optima under perturbations of parameters in S.

Applications

In Optimization

In optimization, set-valued functions are fundamental for addressing nonsmooth and variational problems, where single-valued derivatives may fail to capture the necessary information. A primary application is the subdifferential of a convex function f: \mathbb{R}^n \to \overline{\mathbb{R}}, defined as the set \partial f(x) = \{ v \in \mathbb{R}^n \mid f(y) \geq f(x) + \langle v, y - x \rangle \ \forall y \in \mathbb{R}^n \}, which generalizes the gradient to enable optimization of nondifferentiable objectives while preserving convexity properties like monotonicity of the operator \partial f. This set-valued structure allows the formulation of necessary and sufficient optimality conditions, such as $0 \in \partial f(x^*) for minimizers x^* of convex f, facilitating the analysis of problems like linear programming with piecewise linear costs. Set-valued functions also underpin variational inequalities, which model equilibrium and constrained optimization tasks. A set-valued variational inequality seeks x \in K such that \langle y - x, v \rangle \geq 0 for all y \in K and v \in F(x), where K is a convex set and F: K \rightrightarrows \mathbb{R}^n is a monotone set-valued mapping.^[23] Existence of solutions is guaranteed under hemicontinuity and monotonicity of F via the Browder-Hartman-Stampacchia theorem, extended to multi-valued operators, which ensures solvability in reflexive Banach spaces and applies to problems like traffic flow or contact mechanics.^[23] This framework unifies single-valued and set-valued cases, with the inclusion $0 \in F(x) representing a special zero-finding problem solvable when F is maximal monotone. The epigraphical approach leverages set-valued functions for multiobjective or vector optimization by considering the epigraph \operatorname{epi} F = \{ (x, y) \mid y \in F(x) \}, transforming minimization of F into scalarized problems over this set-valued representation.^[24] Scalarization techniques, such as linear functionals, reduce the task to finding points in \operatorname{epi} F that minimize weighted objectives, enabling the derivation of Pareto-efficient solutions without assuming total order on the range space.^[24] This method is particularly effective for robust optimization, where uncertainty in objectives yields set-valued mappings. For computational resolution, proximal point algorithms address inclusions $0 \in F(x), where F is a maximal monotone set-valued operator, by iterating x^{k+1} \in (I + \lambda_k F)^{-1}(x^k) with \lambda_k > 0, converging weakly to a zero of F in Hilbert spaces. Extensions incorporate variable metrics for faster convergence in nonsmooth settings, such as bundle methods for composite objectives.^[25] Existence of solutions in these algorithms often relies briefly on selection theorems ensuring single-valued resolvents from continuous selections of F.

In Economics and Control Theory

In economics, set-valued functions, often termed correspondences, play a central role in modeling consumer demand and market equilibria. The Walrasian correspondence, for instance, maps prices to the set of optimal consumption bundles that maximize a consumer's utility subject to a budget constraint, formally defined as W(p) = \{ x \mid x \text{ solves } \max u(x) \text{ s.t. } p \cdot x \leq w \}, where p denotes prices, w is wealth, u is the utility function, and x is the consumption bundle.^[26] This set-valued formulation accounts for potential multiplicity of optima under non-strict preferences or indifference curves, enabling the analysis of aggregate excess demand functions in general equilibrium models. Equilibrium existence is established by showing that a price vector equates aggregate demand to supply, often via fixed-point theorems applied to these correspondences.^[26] The use of set-valued functions in economics traces back to the mid-20th century, particularly in Gérard Debreu's axiomatic framework, which extended the Arrow-Debreu model by incorporating correspondences to handle non-convexities and ensure equilibrium under broader conditions.^[27] In this context, Debreu's work in the 1950s formalized how set-valued demand ensures the continuity and upper hemicontinuity properties necessary for equilibrium proofs, influencing modern general equilibrium theory.^[27] Berge's maximum theorem, which guarantees the upper hemicontinuity of the argmax correspondence under continuous objectives and compact constraints, underpins these applications in economic modeling.^[28] In game theory, set-valued functions model strategy sets S_i for player i as correspondences, allowing for constraints or mixed strategies, while payoffs become set-valued under uncertainty to represent ambiguity or imprecise outcomes.^[29] For example, in games with payoff uncertainty, the payoff function for each player maps strategy profiles to sets of possible utilities, capturing scenarios where outcomes depend on unresolved parameters like opponent types or environmental factors; equilibria are then defined as strategy profiles where each player's selection is optimal against the sets induced by others.^[29] This approach extends Nash equilibrium to robust solutions, ensuring stability against perturbations in uncertain environments. In control theory, set-valued functions describe differential inclusions, where the dynamics are given by \dot{x} \in F(x, u) with u \in U, modeling systems with multiple possible velocities or uncertain controls.^[30] Reachability sets, defined as R(t) = \{ x(t) \mid \dot{x} \in F(x, u), \, u \in U, \, x(0) \in X_0 \}, quantify all states attainable from an initial set X_0 over time t, essential for analyzing controllability in nonlinear or hybrid systems.^[30] Viability theory, developed through this framework, ensures the existence of controls that keep trajectories within a viable set, preventing escape from constrained regions like safe operating bounds in engineering applications.^[30]

References

[1]
[PDF] Introduction to Correspondences - P.J. Healy
In order to capture this dependence we would like to have a notion of a set-valued function. The seemingly obvious idea of a function f : X → 2Y from the ...
[2]
[PDF] 6 Lecture 6
A correspondance is just a set-valued function: a correspondance from. X to Y is a map that takes every element in X and maps it to a non-empty subset of Y ...
[3]
[PDF] Non-continuous set-valued functions
Definition 1; [1] For any sets X and Y, F: X - 2Y is a set-valued ... Let F: (X, S) •» 2Y be a set-valued function defined by. Page 16. F(x) □ {f(x) ...
[4]
[PDF] Fixed Point Theorems
Kakutani [6]. In 1941, Kakutani extended the theorem to set-valued functions Φ from a set to its power set: Theorem 2 (Kakutani 1941). Let Φ be an upper ...
[5]
[PDF] Set-Valued Maps And Their Applications - Concordia's Spectrum
They are an extension and a generalization of single- valued maps. Their use in applications such as in control theory, economics and management, biology and ...
[6]
[PDF] A Lower Semicontinuous Regularization for Set-Valued Mappings ...
Let us consider a set-valued mapping F : X ⇉ Y , that is a mapping which assigns to each x ∈ X a subset F(x) of Y . When F(x) 6= ∅ we say that x ∈ DomF ...Missing: scholarly | Show results with:scholarly
[7]
[PDF] arXiv:0907.5439v5 [math.OC] 30 Dec 2010
Dec 30, 2010 · We say that the set-valued map S : X ⇒ Y is closed-valued if S(x) is closed for all x ∈ X, and it is convex-valued if S(x) is convex for all x ∈ ...Missing: scholarly | Show results with:scholarly
[8]
[PDF] A concept of inner prederivative for set-valued mappings ... - Numdam
Assume that there is a neighborhood U of ¯x such that F is closed-valued and convex-valued ... set-valued mapping theorem we need the following definitions and ...Missing: scholarly | Show results with:scholarly
[9]
[PDF] Pointwise Topological Convergence and Topological Graph ...
Then Gr(f) is the graph of a lower quasicontinuous set-valued map. The above Lemma in conjuction with Theorem 2.1 below show that every minimal usco map with ...Missing: scholarly | Show results with:scholarly
[10]
[PDF] ENGLISH FOR MATHEMATICIANS
multivalued function (multifunction, many-valued function, set-valued function, set-valued map, point-to-set map, multi-valued map, multimap) – многозначная ...
[11]
(PDF) An Inverse Mapping Theorem for Set-Valued Maps
Aug 6, 2025 · We prove that certain Lipschitz properties of the inverse F-1 of a set-valued map F are inherited by the map (f + F)-1 when f has vanishing ...Missing: scholarly | Show results with:scholarly
[12]
[PDF] Gerard Debreu on the Integration of Correspondences
Sep 12, 2005 · Segal-Kunze are focussed on the integral of a function rather than a correspondence and they do not have mathematical economics in mind.Missing: origin | Show results with:origin
[13]
Definition of correspondence - Mathematics Stack Exchange
Apr 28, 2013 · Correspondences come up routinely in mathematical economics. One generally calls a map f:X→2Y a 'correspondence' or a set-valued map. These ...Missing: origin | Show results with:origin
[14]
https://baylor-ir.tdl.org/server/api/core/bitstreams/34149edc-3f2d-4016-b2a3-ab1720848a11/content
[15]
multi-valued function in nLab
May 26, 2023 · In older literature (into the 20th century, especially in analysis), functions were often considered to be multi-valued by default, requiring ...
[16]
[PDF] History of Riemann surfaces
Oct 11, 2005 · The problem with multivalued functions still existed however, and at the time of the introduction of complex functions, a 25 year old ...
[17]
https://users.renyi.hu/~descript/papers/Michael_1.pdf
[18]
[PDF] Numerical aspects of the sweeping process - HAL
Jul 29, 2016 · Jean Jacques Moreau. Numerical aspects of the sweeping process ... set-valued mapping). It is assumed, for every t in 1, that C(t) is a ...
[19]
Set-Valued Analysis - SpringerLink
The book under review gives a comprehensive treatment of basically everything in mathematics that can be named multivalued/set-valued analysis.
[20]
https://baylor-ir.tdl.org/server/api/core/bitstrea...
... composition of set-valued functions. If X, Y , and Z are sets, F : X → 2Y and G : Y → 2Z , then we define G ◦ F : X → 2Z by G ◦ F (x) = ⋃ y∈F (x) G(y) ...<|control11|><|separator|>
[21]
Representing uncertainty on set-valued variables using belief ...
A formalism is proposed for representing uncertain information on set-valued variables using the formalism of belief functions.
[22]
Set-Valued Functions of Two Variables in Economic Theory
An application to the problem of the existence of a random price equilibrium is also given.<|control11|><|separator|>
[23]
[PDF] Continuous Selections. I Ernest Michael The Annals of Mathematics ...
Jun 19, 2007 · To show how this can be done, and how the resulting theorems can be applied, is the purpose of this sequence of papers. Let us now introduce ...
[24]
A generalization of Brouwer's fixed point theorem - Project Euclid
A generalization of Brouwer's fixed point theorem. Shizuo Kakutani. DOWNLOAD PDF + SAVE TO MY LIBRARY. Duke Math. J. 8(3): 457-459 (September 1941).
[25]
[PDF] Introduction to Topological Spaces and Set-Valued Maps (Lecture ...
Aug 25, 2006 · Definition 5.1.1 (set-valued map). Let X and Y be topological ... Suppose F(·) a closed valued closed map. Assume there is x0 ∈ X ...Missing: scholarly | Show results with:scholarly
[26]
[PDF] VARIATIONAL ANALYSIS - UW Math Department
Variational analysis is a mathematical theory related to optimization, equilibrium, control, and stability of linear and nonlinear systems.
[27]
Fractional Reverse Inequalities Involving Generic Interval-Valued ...
An interval-valued function is a special type of set-valued function with co-domain as the collection of all real positive intervals. Moore was the first ...<|control11|><|separator|>
[28]
Continuous Selections. I - jstor
ANNALS OF MATHEMATICS. Vol. 63, No. 2, March, 1956. Printed in U.S.A.. CONTINUOUS SELECTIONS. I. BY ERNEST MICHAEL'. (Received December 9, 1954). (Revised May 6 ...Missing: PDF | Show results with:PDF
[29]
Browder-Hartman-Stampacchia variational inequalities for multi ...
An existence theorem of Browder-Hartman-Stampacchia variational inequalities is extended to multi-valued monotone operators.
[30]
Epigraphical Analysis - ScienceDirect.com
As indicated earlier, the epigraphical approach consists in deriving the properties of those maps from those of the multifunction i.e., an epigraph-valued ...
[31]
[PDF] The variable metric proximal point algorithm for monotone operators
Abstract. The proximal point algorithm (PPA) is a method for solving inclusions of the form. 0 ∈ T(z), where T is a monotone operator on a Hilbert space.
[32]
[PDF] Existence of an Equilibrium for a Competitive Economy Kenneth J ...
Oct 9, 2007 · The main results of this paper are two theorems stating very general conditions under which a competitive equilibrium will exist. Loosely ...<|control11|><|separator|>
[33]
[PDF] THEORY OF VALUE An Axiomatic Analysis Of Economic Equilibrium
The two central problems of the theory that this monograph pre- sents are (1) the explanation of tbe prices of commodities resulting from the interaction of ...
[34]
[PDF] A GENERALIZED THEOREM OF THE MAXIMUM
Clearly, the hypotheses of Berge's theorem are not satisfied. Nevertheless, there is no fundamental reason why the Theorem of the Maximum should fail. Indeed, ...
[35]
Towards a Theory of Games with Payoffs that are Probability ... - arXiv
Jun 24, 2015 · This work introduces a new kind of game in which uncertainty applies to the payoff functions rather than the player's actions.
[36]
Differential Inclusions: Set-Valued Maps and Viability Theory
Book Title: Differential Inclusions · Book Subtitle: Set-Valued Maps and Viability Theory · Authors: Jean-Pierre Aubin, Arrigo Cellina · Series Title: Grundlehren ...