InformationTheory

InformationTheory.Shannon.FisherDeBruijnGaussian

source

Fully-internal Gaussian de Bruijn witness #

Assembles the Gaussian heat-kernel derivative facts into a de Bruijn heat-flow witness.

Both analytic halves of ∂_t g_t = (1/2) Δ_x g_t for the centred Gaussian heat kernel g_t = heatKernel t are proved internally:

  • spatial: isHeatSpatialDerivHyp_gaussian (FisherInfo/HeatFlow.lean)
  • time: isHeatTimeDerivHyp_gaussian (GaussianPDFVarianceDerivative.lean)

Main statements #

  • isHeatFlowConvolutionHyp_heatKernel — convolution sub-predicate (positivity + measurability internal; only Z_law remains as input).
  • isHeatFlowDensity_gaussian_heatKernelIsHeatFlowDensity witness with the heat equation backed by isHeatTimeDerivHyp_gaussian.
  • heatKernel_heat_equation∂_t g_t = (1/2) Δ_x g_t with Δ_x g_t = spatialLaplacianHeatKernel.
  • deBruijn_gaussian_heatFlow_witness — de Bruijn identity with heat-flow-density side internal.
  • isRegularDeBruijnHypV2_gaussian_heatFlow — packaged IsRegularDeBruijnHypV2.

Implementation notes #

Remaining open hypotheses consumed as arguments:

  • Z_law : P.map Z = gaussianReal 0 1 — definitional hypothesis that Z is standard normal.
  • IsIBPHypothesis — integration-by-parts / dominated-convergence step (Cover-Thomas).

Convolution sub-predicate #

IsHeatFlowDensity witness (heat equation fully internal) #

The heat equation, with both halves proven #

Final glue — Gaussian de Bruijn witness #