Skip to content

Approximation Theory

← Back to Domain-Specific Abstractions by Domain

6 domain-specific abstractions whose origin domain is Approximation Theory.

  • Barron space — A function space characterized by integral representations or spectral moment bounds that control approximation by two-layer neural networks with dimension-favorable error rates.
  • Lebesgue's lemma — An approximation bound stating that a bounded linear projection's error is at most one plus its operator norm times the best attainable error from the target subspace.
  • Perfect spline — A univariate spline of order m whose m-th derivative takes alternating values plus or minus one between successive knots.
  • Ridge function — Factor a multivariate function through one linear or affine projection, so its value varies only along a selected direction and remains constant across every orthogonal affine slice.
  • Variation diminishing property — A transformation is variation diminishing when it cannot increase sign changes, oscillations, or intersection counts, making its output no more geometrically or algebraically variable than its input control data.
  • Zolotarev polynomials — Extremal polynomials with prescribed leading coefficients that minimize uniform deviation on an interval, generalizing Chebyshev polynomials in approximation theory.