theorem
proved
finite_resolution_cell_finite_events
show as:
view math explainer →
open explainer
Generate a durable explainer page for this declaration.
open lean source
IndisputableMonolith.RecogGeom.FiniteResolution on GitHub at line 63.
browse module
All declarations in this module, on Recognition.
explainer page
depends on
formal source
60
61/-- If R has finite local resolution at c, the resolution cell at c
62 has a finite number of "neighbors" in any finite-resolution neighborhood -/
63theorem finite_resolution_cell_finite_events (c : C)
64 (h : HasFiniteLocalResolution L r c) :
65 ∃ U ∈ L.N c, ∀ c' ∈ U, r.R c' ∈ r.R '' U ∧ (r.R '' U).Finite := by
66 obtain ⟨U, hU, hfin⟩ := h
67 use U, hU
68 intro c' hc'
69 exact ⟨⟨c', hc', rfl⟩, hfin⟩
70
71/-! ## Discrete Local Recognition Geometry -/
72
73/-- A recognition geometry is locally discrete if events are finite everywhere -/
74def IsLocallyDiscrete (L : LocalConfigSpace C) (r : Recognizer C E) : Prop :=
75 HasFiniteResolution L r
76
77/-- In a locally discrete recognition geometry, every neighborhood contains
78 only finitely many distinguishable configurations -/
79theorem locally_discrete_finite_classes
80 (h : IsLocallyDiscrete L r) (c : C) :
81 ∃ U ∈ L.N c, (r.R '' U).Finite :=
82 h c
83
84/-! ## No Continuous Injection Theorem -/
85
86/-- **Key Insight**: If a neighborhood has infinite configurations but finite
87 events, then the recognizer cannot be injective on that neighborhood.
88
89 This explains why discrete recognition geometry fundamentally differs
90 from continuous Euclidean geometry. -/
91theorem no_injection_on_infinite_finite (c : C)
92 (U : Set C) (hU : U ∈ L.N c)
93 (hinf : Set.Infinite U) (hfin : (r.R '' U).Finite) :