Epigraph¶
The upward-closed set of real-height pairs above an extended-real-valued function, preserving each function value as a fiber infimum.
Core Idea¶
For \(f:X\to\mathbb R\cup\{+\infty\}\), its epigraph is \(\{(x,t)\in X\times\mathbb R:t\geq f(x)\}\). Each finite value is the lower endpoint of an upward vertical ray; at \(f(x)=+\infty\) the real-height fiber is empty, and \(\inf\varnothing=+\infty\) recovers the value. The set represents the function without requiring it to be convex, continuous or optimized.[ref-ad86641c4e3b][ref-43644f29a18c]
Scope of Application¶
In convex analysis, a Euclidean function is convex exactly when its epigraph is convex; under Nemirovski's stated extended-real Euclidean convention, lower semicontinuity corresponds to closed epigraph. In optimization, a max-of-affine objective has epigraph constraints \(a_i^Tx+b_i\leq t\), so minimizing \(t\) is a linear program for that specific form. A general epigraph lift makes the objective linear but need not make the constraints linear.[ref-ad86641c4e3b][ref-43644f29a18c]
Clarity¶
The graph retains only equality pairs \((x,f(x))\); the hypograph contains heights below the value. An epigraph requires every real height at or above it. Convexity, closedness and finite solver representation are separate properties, not constitutive roles.[ref-ad86641c4e3b][ref-43644f29a18c]
Manages Complexity¶
The construction turns value comparisons into set membership while preserving every function value as a fiber infimum. For a nonempty family on a common domain, the epigraph of the pointwise supremum is the intersection of the epigraphs. This can simplify a proof, but the function/domain/topology assumptions still govern which conclusion is valid.[^ref-43644f29a18c]
Abstract Reasoning¶
At each input, check that the proposed section is empty or a complete inclusive upper ray. Recover the lower endpoint and test whether it equals the claimed \(f(x)\). Then ask separately whether the set is convex or closed and whether its membership inequalities fit the intended optimization method.[ref-ad86641c4e3b][ref-43644f29a18c]
Knowledge Transfer¶
The same upward-fiber structure serves the \(x^2\) geometry example and the max-of-affine optimization lift. Live Function (Mapping) is a proposed presupposed skeleton because the rays need values \(f(x)\); Graph of a Function and Hypograph are distinct neighboring set constructions, not parents.[ref-ad86641c4e3b][ref-43644f29a18c]
[^ref-ad86641c4e3b]: Stephen Boyd and Lieven Vandenberghe, Convex Optimization lecture slides, author-hosted PDF with revised slides credited to Boyd, Vandenberghe and Parth Nobel, slide 3.12/PDF p. 56 and slide 4.15/PDF p. 102. [^ref-43644f29a18c]: Aharon Ben-Tal and Arkadi Nemirovski, Optimization III: Convex Analysis, Nonlinear Programming Theory, Nonlinear Programming Algorithms, Georgia Tech lecture notes (Fall 2020), Lecture 2 §2.6.1, PDF pp. 72–74, Proposition 2.6.1 and Corollary 2.6.1.
Relationships to Other Abstractions¶
Current abstraction Epigraph Domain-specific
Parents (1) — more general patterns this builds on
-
Epigraph presupposes Function (Mapping) Prime
Each upward fiber is bounded by a value f(x) assigned by a function.
Hierarchy path (1) — routes to 1 parentless root
- Epigraph → Function (Mapping)
Neighborhood in Abstraction Space¶
Epigraph sits in a sparse region of the domain-specific corpus (72nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Graph Structures & Combinatorial Objects (44 abstractions)
Nearest neighbors
- Slice Sampling — 0.85
- Norm — 0.84
- Euclidean Space — 0.83
- Normal Order of an Arithmetic Function — 0.83
- Invex Function — 0.83
Computed from structural-signature embeddings · 2026-10-08