Graph of a Function¶

Represent a function by the set of ordered input–output pairs selected by its evaluation rule, preserving every domain element with exactly one associated value while separating the graph from a plotted picture or graph-theoretic network.

Version: v2 · 2026-09-06 · History
Domain-specific #: 1955
Origin domain: mathematics
Subdomain: foundations and analysis
Aliases: Function graph, Graph of f

Core Idea¶

For a function f: X -> Y, its graph is the set Graph(f) = {(x,f(x)) : x in X}, viewed as a subset of the Cartesian product X x Y. The graph turns an evaluation rule into a relational extension: every domain element appears as a first coordinate, and functionality requires exactly one second coordinate for it. MIT's introductory analysis text uses this input–output understanding to connect functions with their graphical representation, while standard set-theoretic treatments identify a function with a special relation.^[1] The abstraction is the ordered-pair set, not merely ink on coordinate axes.

The graph can serve as a definition of a function when domain and codomain conventions are supplied. A relation G subseteq X x Y is the graph of a total function on X exactly when for every x in X there exists a unique y in Y with (x,y) in G. Existence gives totality; uniqueness gives single-valuedness. If existence is required only for some inputs, the relation represents a partial function. If uniqueness fails, it is not the graph of a single-valued function under that orientation, even if a geometric curve can be drawn.

The ordered-pair graph preserves the domain and realized image but does not, by itself, preserve unused elements of a declared codomain. The same pair set can represent a function X -> Y and a function X -> Z when both codomains contain the realized values. This is why some foundations identify a function with a triple containing domain, codomain, and graph, while others identify it with the graph under ambient conventions. Halmos's relational treatment makes the set-of-pairs identity explicit and supports this distinction.^[2] A reference entry must say which convention is in force.

For real-valued functions on subsets of the real line, the abstract graph becomes a planar subset and can be plotted. The vertical-line test is then the uniqueness condition rendered geometrically: a vertical line over an input meets the graph at most once, and exactly once if the function is total on that input. This special picture is pedagogically powerful but not the general definition. Functions between matrices, groups, strings, manifolds, or infinite-dimensional spaces have graphs even when no literal two-dimensional plot is available.

Graphs support structural reasoning. Equality of functions with fixed domain and codomain reduces to equality of graph pairs. The graph of a composite is obtained through an existential intermediate value rather than naive set composition. The graph of an injective function has a converse relation that remains single-valued on the image; the graph of a bijection can be transposed to give the inverse function. In topology and analysis, graph closure, closed-graph conditions, measurability, and smooth embedded-graph constructions turn properties of a mapping into properties of a subset of a product space. The candidate is therefore more than a diagram convention and is not covered by graph-theoretic network nodes.

Structural Signature¶

The function. A mapping assigns each input in a declared domain exactly one output.
The domain. Every domain element must occur as a first coordinate for a total function.
The codomain. An ambient output set bounds the second coordinates but may contain unused values.
The Cartesian product. Ordered pairs live in X x Y, preserving coordinate roles.
The evaluation pair. Each graph element has the form (x,f(x)).
The existence condition. At least one pair occurs for each domain input.
The uniqueness condition. At most one output is paired with each input.
The realized image. Projection onto the second coordinate recovers the function's image.
The orientation. Swapping coordinates produces a converse relation, not automatically a function.
The ambient representation. A plot, table, database relation, or subset presentation may display the same pair structure.

What It Is Not¶

Not a graph-theory graph. Vertices and edges are not the defining objects here.
Not merely a plotted curve. The ordered-pair set exists without coordinates or a drawing.
Not the image of the function. The image forgets which input produced each output.
Not the codomain. The codomain can contain values absent from the graph's second projection.
Not an arbitrary binary relation. Totality and single-valuedness impose exact conditions.
Not an epigraph. An epigraph contains points above scalar function values rather than only equality pairs.
Not automatically invertible. The converse is a function only under appropriate injectivity and domain restrictions.

Scope of Application¶

The graph-of-a-function abstraction is literal whenever a mapping is represented by its complete set of ordered evaluation pairs in a declared product.

Set-theoretic foundations. Defining functions as special binary relations.
Elementary analysis. Plotting real functions and testing single-valuedness.
Topology. Studying closed graphs and embeddings into product spaces.
Functional analysis. Representing operators through graph subspaces and domains.
Algebra. Encoding homomorphisms as subsets of product structures.
Category theory. Comparing graph relations with morphisms under extra structure.
Databases. Storing finite function tables as key–value relations.
Formal verification. Reducing evaluation and extensional equality to pair membership.

Clarity¶

A clear statement writes f: X -> Y, defines Graph(f) as a subset of X x Y, and distinguishes domain, image, and codomain. If a relation is offered first, it checks both existence and uniqueness. It declares whether the foundational convention identifies a function with the pair set or retains domain and codomain as extra data. In a real plot, it identifies axes and excludes inputs outside the domain; a curve segment or sampled plot is not silently equated with the exact graph. For a partial function it states the actual first-coordinate projection. Claims about inverse functions state injectivity or restrict the converse to the realized image.

Manages Complexity¶

The graph externalizes a mapping's behavior as a static object. Evaluation becomes pair lookup, equality becomes extensional pair equality, domain and image become coordinate projections, restriction becomes intersection with a smaller input product, and inversion becomes coordinate exchange plus a functionality test. This lets set, topological, algebraic, and analytical tools act on mappings. The compression has a cost: a finite picture may omit resolution, the pair set may omit declared but unused codomain values, and a high-dimensional graph can be less intuitive than an evaluation rule. Reference-grade use therefore keeps the ambient product and function convention explicit.

Abstract Reasoning¶

Declare the domain, codomain, and evaluation rule of the function.
Form an ordered pair from every input and its unique output.
Collect the pairs as a subset of the Cartesian product.
Project first coordinates to verify total domain coverage.
Test uniqueness of second coordinates at each fixed first coordinate.
Project second coordinates to recover the realized image.
Restrict or extend the domain by changing which first coordinates are represented.
Swap coordinates only after checking whether the converse remains functional.
Apply subset properties such as closure, measurability, smoothness, or algebraic closure in the product.
Distinguish the exact pair set from any finite table, sample, or plotted rendering.

Knowledge Transfer¶

The graph transfers the general move from an operation to its extension: replace a rule by the relation containing every input–output instance. The move supports database key constraints, state-transition relations, operator graphs, and relational semantics. Functionality is the transferable diagnostic: one output per declared input. The domain accent is ordered pairs in a Cartesian product and the exact mathematical mapping convention. Network terminology does not transfer, and a graph-theoretic edge set should not be imported merely because the same English word appears.

Examples¶

Canonical¶

Let f: R -> R be f(x)=x^2. Its graph is {(x,x^2): x in R}. The inputs 2 and -2 produce distinct pairs with the same second coordinate, so the function is not injective, but every vertical line at an allowed input meets the graph once. The second-coordinate projection is the nonnegative real axis, not the whole declared codomain. Reflecting the parabola across the diagonal produces a relation that is not a function on the nonnegative reals unless one branch is selected.^[1]

Mapped back: evaluation rule x squared → complete ordered-pair set → vertical uniqueness → image projection → converse fails without branch restriction.

Applied / In Practice¶

Let T be a linear operator with a specified domain in a normed space. Its graph is the linear subspace {(x,Tx)} of the product space. A sequence of pairs can converge in the product even when the operator is unbounded; asking whether the limit pair remains in the graph produces the closed-graph condition. The reasoning uses the same pair-set abstraction as the parabola, but a two-dimensional plot is irrelevant. Standard topology texts treat products and graphs as subset constructions supporting such closure questions.^[3]

Mapped back: operator evaluation → graph subspace in a product → convergence of pairs → closedness test → operator-level conclusion.

Structural Tensions¶

Rule vs. extension. A formula is compact while a graph lists all cases abstractly. Diagnostic: Can evaluation be recovered uniquely from pair membership?
Graph vs. plot. Pictures aid intuition but are sampled and coordinate-bound. Diagnostic: Would the definition still work for a function between groups?
Image vs. codomain. Pair projection records used outputs only. Diagnostic: Are unused codomain elements part of function identity under the chosen convention?
Relation vs. function. Any pair set is a relation, but functionality is restrictive. Diagnostic: Does each domain input have exactly one partner?
Converse vs. inverse. Coordinate swap is always possible but may not be functional. Diagnostic: Is the original mapping injective on the stated domain?
Finite evidence vs. exact graph. Tables and plots show only selected values. Diagnostic: What justifies extrapolation to every domain point?
Static subset vs. operational evaluation. The graph stores extension while an algorithm computes values. Diagnostic: Is complexity of computation being confused with existence of pairs?

Structural–Framed Character¶

The structure is function, domain coverage, unique output, ordered pair, Cartesian product, image projection, and converse relation. The frame is the chosen coding of ordered pairs, declared codomain convention, coordinate system, drawing, table, or software representation. A different plot can preserve the graph; a pair relation with two outputs for one input does not.

Structural Core vs. Domain Accent¶

The transferable core is rule assigning outputs → complete set of input–output instances → relational tests and projections. The domain accent is function extensionality, ordered pairs, Cartesian products, vertical-line visualization, inverse relations, and closed graphs. Remove the accent and Function Mapping remains; retain it and Graph of a Function is a distinct foundational construction.

Function Mapping is the strict parent by composition. The graph materializes every assignment made by a function as an ordered pair and preserves the mapping's domain-to-output relation. Function Mapping is broader and need not be represented as a subset of a Cartesian product.

The prospective workspace queue contains one strict upward edge to prime:function_mapping. No live DAG mutation is authorized.

Relationships to Other Abstractions¶

Current abstraction Graph of a Function Domain-specific

Parents (1) — more general patterns this builds on

Graph of a Function is a kind of Function (Mapping) Prime

Function Mapping is the strict parent by composition.

Hierarchy path (1) — routes to 1 parentless root

Graph of a Function → Function (Mapping)

Neighborhood in Abstraction Space¶

Graph of a Function sits in a sparse region of the domain-specific corpus (82^nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Functions, Maps & Integral Structure (10 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Graph-theory graph. A vertex–edge structure.
Plot. A visual rendering, often sampled and coordinate dependent.
Image of a function. The set of attained outputs without input associations.
Binary relation. A broader pair set that need not be total or single-valued.
Epigraph. Points satisfying an inequality above a real function.
Level set. Inputs sharing one output value.
Inverse relation. The coordinate-swapped graph, not always an inverse function.

References¶

[1] MIT OpenCourseWare, 18.013A Calculus with Applications, section 1.3, Functions, https://ocw.mit.edu/ans7870/18/18.013a/textbook/HTML/chapter01/section03.html. registry ↩a ↩b

[2] Paul R. Halmos, Naive Set Theory (Springer, 1974), chapters on relations and functions, https://doi.org/10.1007/978-1-4757-1645-0. registry ↩

[3] James R. Munkres, Topology, 2^nd ed. (Prentice Hall, 2000), product spaces, continuity, and graph exercises, ISBN 978-0-13-181629-9. registry ↩