Error propagation through an approximate linear map (Lemma 2.7)
ProvedIntMul.HvdH.error_propagationinteger-multiplication
Let be a continuous -linear map between normed spaces with , and let be any function. If and , then
This is Lemma 2.7 of Harvey–van der Hoeven (errors measured in units of there).
Preamble
import Mathlib
Formal statement
namespace IntMul.HvdH
theorem error_propagation {V W : Type*} [NormedAddCommGroup V] [NormedSpace ℂ V] [NormedAddCommGroup W] [NormedSpace ℂ W]
(A : V →L[ℂ] W) (hA : ‖A‖ ≤ 1) (Aa : V → W) (v vt : V) (eA ev : ℝ)
(hAa : ‖Aa vt - A vt‖ ≤ eA) (hv : ‖vt - v‖ ≤ ev) : ‖Aa vt - A v‖ ≤ eA + ev := by sorry
end IntMul.HvdHSource
Harvey, van der Hoeven, Integer multiplication in time O(n log n), Annals of Math. 193 (2021), https://hal.science/hal-02070778, Section 2.6