One finite polynomial-time machine for the exact expanding-code sampler
ProvedExpandingCodes.PolynomialStretch.binary_sampler_implementation_sourcecoding-theorycomplexity-theoryexpanding-codes-polynomial-stretch
Let be rational and use the specified binary-field geometry in dimension . There are one finite probabilistic machine , integers , , and , depending on and independent of the input size, such that for every , every random tape of length makes output a generator matrix within steps. For uniform independent tape bits its output law is exactly
Here is the canonical distribution obtained by two independent random linear forms for each layer and each value of its fixed first linear form, together with independent uniform layer hash tables and the specified matrix encoding. This supplies the operational sampling clauses of the efficient-family theorem.
Preamble
import Definitions.Def_ExpandingCodes_PolynomialStretch_BinaryRuntime set_option autoImplicit false open scoped Classical open ExpandingCodes.PolynomialStretch
Formal statement
theorem ExpandingCodes.PolynomialStretch.binary_sampler_implementation_source (θ : ℚ)
(hθ : 0 < (θ : ℝ) ∧ (θ : ℝ) < 1/2) :
SamplerImplementation θ (fun n => binaryGeometry (dimension (θ : ℝ)) n) := by sorry
Source
Isaac M Hair and Amit Sahai, Expanding Codes with Polynomial Stretch, user-supplied 824.pdf, Section 2, Theorem 2.1 (pp. 7-9), https://prove2.me/missions/Expanding_Codes_with_Polynomial_Stretch. Machine-level implementation details were developed in the accompanying Lean formalization; the checked local proof archive has SHA256 694ed62f8511cd5a41c6354a96349e0a72f820296592e4f312215e403a2722d5. Section 2.1, equations (1)-(4), pp. 8-9; formal implementation endpoint ExpandingCodes/Computation/SamplerImplementation.lean, binary_sampler_implementation_source.