Polynomial map for the linear GUV condenser #
The source polynomial is repeatedly squared modulo a supplied polynomial.
Evaluating every rth iterate gives the Parvaresh--Vardy neighbor map with
exponent h = 2^r, omitting the seed coordinate. This is the map in
Guruswami, Umans, and Vadhan (2009), Section 3.1, with the characteristic-power
choice used by Cheraghchi (2010), Section 2.3.3, to obtain source linearity.
These definitions use Mathlib's noncomputable polynomial representation.
They specify the algebraic map and its modular iteration. Condenser.Encoding
and its correctness layer implement this map for the explicit binary fields;
the expansion and flat lossless bounds are proved in separate condenser modules.
No basis or irreducible polynomial is chosen in these general definitions.
For a nonmonic modulus, Mathlib's %ₘ leaves the dividend unchanged; the
degree guarantee requires monicity, and the eventual expansion proof also
needs the construction's irreducibility hypotheses.
Repeated modular squaring, starting with the reduced source polynomial.
For monic E, every stored polynomial has degree less than E.degree.
Equations
- Algebraic.Cutwidth.Extractor.modularFrobenius E 0 x✝ = x✝ %ₘ E
- Algebraic.Cutwidth.Extractor.modularFrobenius E j.succ x✝ = Algebraic.Cutwidth.Extractor.modularFrobenius E j x✝ ^ 2 %ₘ E
Instances For
The polynomial output coordinates of the GUV map for exponent 2^r.
The seed y is not included in the output, matching the strong condenser
convention. On source polynomials of degree less than a monic E, coordinate
zero is f.eval y.
Equations
- Algebraic.Cutwidth.Extractor.polynomialCondenser E r m f y i = Polynomial.eval y (Algebraic.Cutwidth.Extractor.modularFrobenius E (r * ↑i) f)