def
definition
is_complete_correction
show as:
view math explainer →
open explainer
Read the cached plain-language explainer.
open lean source
IndisputableMonolith.Thermodynamics.ErrorCorrection on GitHub at line 84.
browse module
All declarations in this module, on Recognition.
explainer page
depends on
formal source
81
82/-- A correction protocol is complete if it maps all states to ground states
83 within a finite number of applications. -/
84def is_complete_correction {X : Ω → ℝ} (C : CorrectionProtocol X) : Prop :=
85 ∀ ω, ∃ n : ℕ, Jcost (X (C.correct^[n] ω)) = 0
86
87/-! ## Fault Tolerance Threshold -/
88
89/-- The fault tolerance threshold: below this defect density,
90 errors can be corrected faster than they accumulate. -/
91structure FaultToleranceThreshold where
92 /-- The threshold value -/
93 threshold : ℝ
94 /-- The threshold is positive -/
95 threshold_pos : 0 < threshold
96
97/-- A system is fault-tolerant if defect accumulation is bounded. -/
98def is_fault_tolerant (sys : RecognitionSystem) (X : Ω → ℝ)
99 (ft : FaultToleranceThreshold) : Prop :=
100 ∀ (p : ProbabilityDistribution Ω),
101 expected_cost p.p X < ft.threshold →
102 -- The system can be corrected back to near-equilibrium
103 ∃ (C : CorrectionProtocol X),
104 expected_cost (fun ω => p.p (C.correct ω)) X < expected_cost p.p X / 2
105
106/-! ## Code Distance and φ-Ladder -/
107
108/-- The code distance is the minimum cost of a non-trivial error.
109 In RS, this is related to the φ-ladder structure. -/
110noncomputable def code_distance (X : Ω → ℝ) : ℝ :=
111 ⨅ ω : { ω : Ω // Jcost (X ω) > 0 }, Jcost (X ω.val)
112
113/-- The φ-code distance: minimum cost at a φ-ladder step.
114 d_φ = J(φ) = (√5 - 2)/2 ≈ 0.118 -/