AFTD Auto-Formalizing Theoretical Domains

petz_power_divergence

def

The Petz power divergence petz_power_divergence β ρ σ of type ℝ for an exponent β : ℝ and real matrices ρ, σ : Matrix n n ℝ on a finite type n with decidable equality, defined as (1 - (cfc (fun x : ℝ => x ^ β) ρ * cfc (fun x : ℝ => x ^ (1 - β)) σ).trace) / (β * (1 - β)).

Statement

noncomputable def petz_power_divergence {n : Type*} [Fintype n] [DecidableEq n]
    (β : ℝ) (ρ σ : Matrix n n ℝ) : ℝ
Defined in
OP-1440
Verified
11 Nov 2026
Axioms
none listed
Used by 1
  • petz_power_informationThe Petz power mutual information petz_power_information β ρ of type ℝ for an exponent β : ℝ and a bipartite real matrix ρ : Matrix (u ×…

Read back from the Lean

Given a finite type n with decidable equality, a real number β, and two n × n real matrices ρ and σ, petz_power_divergence β ρ σ is defined as: (1 - tr(cfc (fun x => x ^ β) ρ * cfc (fun x => x ^ (1 - β)) σ)) / (β * (1 - β)) where cfc is the continuous functional calculus, * is matrix multiplication, and trace is the matrix trace.

Written by a model that saw only the Lean, never the English above. If the two disagree, that is worth a challenge.

Lean source view module on GitHub

/-- The Petz power divergence D_beta(rho || sigma) for real matrices. -/
noncomputable def petz_power_divergence {n : Type*} [Fintype n] [DecidableEq n]
    (β : ℝ) (ρ σ : Matrix n n ℝ) : ℝ := (1 - (cfc (fun x : ℝ => x ^ β) ρ * cfc (fun x : ℝ => x ^ (1 - β)) σ).trace) / (β * (1 - β))
See it in the graphDiscuss this resultChallenge it