pith. machine review for the scientific record. sign in
theorem proved term proof high

Jcost_n_zero_iff

show as:
view Lean formalization →

Multi-channel J-cost vanishes precisely when every component equals one. Researchers on the ALEXIS B5 empirical program cite the result to certify that the summed cost reaches its global minimum only at the all-ones vector. The argument unfolds the sum definition, applies the single-channel positivity lemma inside a contradiction for the forward direction, and substitutes the unit value for the converse.

claimLet $n$ be a natural number and let $x$ be a vector in $R^n$ with every component strictly positive. Then the summed cost $J_n(x) = sum_i J(x_i)$ equals zero if and only if $x_i = 1$ for every index $i$.

background

The module defines the multi-channel J-cost as the sum of single-channel J-costs over $n$ independent positive components. This additive extension matches the ALEXIS Exp B5 run, which treated amplitude, phase, and frequency as separate channels whose costs add.

proof idea

Unfold the sum definition of Jcost_n. The forward direction assumes the total sum is zero, isolates any deviant component via the single-summand inequality, and invokes Jcost_pos_of_ne_one to reach a contradiction. The converse replaces each term by zero using Jcost_unit0 and simplifies the sum.

why it matters in Recognition Science

The theorem supplies the zero_iff field of the MultiChannelJCostCert certificate. It completes the uniqueness claim for the multi-channel cost in the ALEXIS B5 formalisation, confirming the all-ones vector as the sole global minimum and supporting the listed descent property. The result aligns with the J-uniqueness step of the Recognition Science forcing chain.

scope and limits

formal statement (Lean)

  48theorem Jcost_n_zero_iff {n : ℕ} (x : Fin n → ℝ) (hx : ∀ i, 0 < x i) :
  49    Jcost_n x hx = 0 ↔ ∀ i, x i = 1 := by

proof body

Term-mode proof.

  50  unfold Jcost_n
  51  constructor
  52  · intro h i
  53    by_contra hi
  54    have hnn : ∀ j : Fin n, 0 ≤ Jcost (x j) := fun j => by
  55      by_cases hj : x j = 1
  56      · rw [hj, Jcost_unit0]
  57      · exact le_of_lt (Jcost_pos_of_ne_one (x j) (hx j) hj)
  58    have hle : Jcost (x i) ≤ ∑ j : Fin n, Jcost (x j) :=
  59      Finset.single_le_sum (fun j _ => hnn j) (Finset.mem_univ i)
  60    linarith [h ▸ hle, Jcost_pos_of_ne_one (x i) (hx i) hi]
  61  · intro hall
  62    have : ∀ i : Fin n, Jcost (x i) = 0 := fun i => by rw [hall i, Jcost_unit0]
  63    simp [this]
  64
  65/-- J_n is symmetric channel-wise. -/

used by (1)

From the project-wide theorem graph. These declarations reference this one in their body.

depends on (10)

Lean names referenced from this declaration's body.