Tensor-product low-degree evaluation codes #
This is the algebraic evaluation code used by the local-recovery gadget in the Boolean mass-production manuscript. Arbitrary data on a finite Cartesian grid is interpolated by a multivariate polynomial and evaluated over the whole ambient affine space.
The semantic node map below is a fixed noncomputable injection obtained from the finite cardinality equivalence. It proves the code theorem independently of representation. A later circuit layer must use the manuscript's concrete polynomial-basis field representation and prove its indexing costs.
The tensor-product Lagrange basis polynomial for one grid point.
Equations
- Algebraic.MassProduction.gridBasis nodes point = ∏ coordinate : Coordinate, (Polynomial.toMvPolynomial coordinate) (Lagrange.basis Finset.univ nodes (point coordinate))
Instances For
Interpolate an arbitrary field-valued message on a Cartesian grid.
Equations
- Algebraic.MassProduction.gridInterpolate nodes message = ∑ point : Coordinate → Index, MvPolynomial.C (message point) * Algebraic.MassProduction.gridBasis nodes point
Instances For
Evaluate the tensor-grid interpolant at an ambient point.
Equations
- Algebraic.MassProduction.evaluationCode nodes message point = (MvPolynomial.eval point) (Algebraic.MassProduction.gridInterpolate nodes message)
Instances For
The paper's grid width floor ((q - 1) / dimension).
Equations
- Algebraic.MassProduction.resourceGridWidth fieldCard dimension = (fieldCard - 1) / dimension
Instances For
A fixed semantic injection of interpolation-node indices into a finite field. This definition is intentionally noncomputable; concrete field indexing belongs to the circuit-realization layer.
Equations
- Algebraic.MassProduction.resourceNodes K dimension point = (Fintype.equivFin K).symm ⟨↑point, ⋯⟩
Instances For
The paper-parameterized systematic resource code.
Equations
- Algebraic.MassProduction.paperEvaluationCode K dimension message point = Algebraic.MassProduction.evaluationCode (Algebraic.MassProduction.resourceNodes K dimension) message point
Instances For
A tensor Lagrange basis polynomial evaluates to one at its own grid point.
A tensor Lagrange basis polynomial vanishes at every other grid point.
The evaluation code agrees with the supplied message at every grid point.
Grid interpolation preserves message addition.
Grid interpolation preserves field scalar multiplication.
Distinct grid messages have distinct ambient evaluation tables.
A tensor basis polynomial has total degree at most the number of coordinates times one less than the grid width.
The complete tensor-grid interpolant obeys the same total-degree budget.
The ambient evaluation table has |K| ^ |Coordinate| coordinates.
The message grid has |Index| ^ |Coordinate| positions.
The fixed semantic node map is injective.
The grid-width choice satisfies the strict affine-line degree bound.
The paper-parameterized code is systematic on its interpolation grid.
Under the exact tensor-degree inequality, every code symbol is the sum of the other symbols on any affine line through it.
The paper's grid-width choice gives affine-line recovery in every positive dimension.