Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Cov comm

Proved
Logic.PhaseRoute.cov_comm

by raver1975 · Sep 12, 2026 · Mathlib c5ea003 (Lean v4.30.0)

aether-cataloglogic

Formal statement of Logic.PhaseRoute.cov_comm from the Aether Catalog (Logic). The mathematical content is given by the Lean statement below; a human-readable write-up is pending.

theorem Logic.PhaseRoute.cov_comm(y h : ι → ℝ) : cov y h = cov h y := by sorry

Formalization Note Transplanted verbatim from the Aether Catalog source Logic/PhaseRouteAlignment.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.

Preamble
-- Thm stub generated from Logic/PhaseRouteAlignment.lean
import Mathlib
import Definitions.Def_Logic_PhaseRouteAlignment
import Definitions.Def_Logic_PhaseRouteLeastSquares
/-
# The phase route is closed, but the interaction route is open — exactly

This file proves a *separation theorem* for feature encodings, in the exact
finite-sample least-squares calculus of `Logic.PhaseRouteLeastSquares`.

Setting.  The sample space is a product `α × β` (think: the residue of a root
position modulo one prime, paired with the residue modulo a second prime, or
paired with itself), and the target is an **alignment indicator**

  `graphInd σ (a, b) = if b = σ a then 1 else 0`   (`σ : α ≃ β`).

*Singleton* (linear-phase) encodings are the additive predictors
`additive u v (a, b) = u a + v b`, where `u`, `v` are **arbitrary** real
functions of a single coordinate — this is strictly more general than one-hot
dummies, sines/cosines of phases, or any other per-coordinate featurisation, and
it also covers arbitrarily many such features used simultaneously, since the
additive family is closed under sums.

Results.

* `cov_graphInd_additive_eq_zero` : every additive predictor has covariance
  *exactly* `0` with the alignment target.  There is no small residual signal to
  be found: the linear phase route is closed identically, not approximately.
* `graphInd_additive_no_gain` / `graphInd_additive_strict_loss` : consequently no
  additive predictor beats the intercept-only baseline, and every *nonconstant*
  one is strictly worse; the excess error equals the predictor's own variance.
* `Rsq_additive_nonpos` : the same statement in `R²` units — the attainable gain
  is `≤ 0`, never the pre-stated `+0.05`, and never even `+0.0215`.
* `nuisance_no_gain` : adding arbitrarily many features from an *independent*
  block of coordinates (other primes / other windows) does not change this.
* `graphInd_eq_sum_interactions` : the alignment target is *exactly* a sum of
  `card α` degree-2 products of one-hot singleton features, and
  `Rsq_interaction_eq_one` : that degree-2 encoding attains `R² = 1`.

So the missing variance is not merely hard to reach with singleton phases; it
lives, provably and entirely, in the degree-2 (joint alignment) layer.
-/

open Logic.PhaseRoute

open Finset

/-! ### Bilinearity tools for the empirical covariance -/

variable {ι : Type*} [Fintype ι] [Nonempty ι]


omit [Nonempty ι] in
Formal statement
theorem Logic.PhaseRoute.cov_comm(y h : ι → ℝ) : cov y h = cov h y := by sorry
Source
https://github.com/paulklemstine/Lean/blob/53c2925a02/Catalog/Logic/PhaseRouteAlignment.lean#L59

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me