Harmonic means and coordinate projections #
Definitions for Section 4 of Oliver Korten, Top-Down Lower Bounds for All Depths, ECCC TR26-221 (2026), https://eccc.weizmann.ac.il/report/2026/221/.
The harmonic mean is zero if any value is zero. This implements the paper's
extended-real convention without using the real-number convention 0⁻¹ = 0
inside the average. Theorems assume nonnegative inputs.
A coordinate set is its Boolean membership function. Projections are represented on the full cube: their value depends only on the selected coordinates. Averaging on the full cube gives each projected pattern the same multiplicity, so this is the uniform average over projected patterns. All expectations are finite sums.
The uniform harmonic mean, with value zero whenever an entry is zero.
Equations
- Complexity.BooleanAnalysis.harmonicMean f = if ∃ (x : α), f x = 0 then 0 else (Finset.univ.expect fun (x : α) => (f x)⁻¹)⁻¹
Instances For
Average over the unselected coordinates, keeping the selected bits of x.
Equations
- Complexity.BooleanAnalysis.projectionAverage f selected x = Finset.univ.expect fun (y : ι → Bool) => f fun (i : ι) => if selected i = true then x i else y i
Instances For
The marginal on actual selected-coordinate patterns, obtained by uniformly
averaging over their completions. For a normalized density f, this is the
normalized density of its coordinate projection.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Korten's harmonic mean transform (Definition 5), represented on the full cube.