Documentation

HighDimProb.Analysis.NetSupremum

theorem HighDimProb.bddAbove_image_of_uniformContinuousOn_of_isInternalEpsilonNet {alpha : Type u_1} [PseudoMetricSpace alpha] {K : Set alpha} {T : Finset alpha} {eps : } {f : alpha} (hT : ∀ (n : ), (T n).Nonempty) (hnet : ∀ (n : ), IsInternalEpsilonNet K (↑(T n)) (eps n)) (heps_pos : ∀ (n : ), 0 < eps n) (heps : Filter.Tendsto eps Filter.atTop (nhds 0)) (hf : UniformContinuousOn f K) :
BddAbove (f '' K)

A shrinking positive internal-net family makes the image of a uniformly continuous real function bounded above. The bound is supplied by one finite net at a scale below the continuity modulus for tolerance 1.

theorem HighDimProb.tendsto_finset_sup'_of_isInternalEpsilonNet {alpha : Type u_1} [PseudoMetricSpace alpha] {K : Set alpha} {T : Finset alpha} {eps : } {f : alpha} (hK : K.Nonempty) (hT : ∀ (n : ), (T n).Nonempty) (hnet : ∀ (n : ), IsInternalEpsilonNet K (↑(T n)) (eps n)) (heps_pos : ∀ (n : ), 0 < eps n) (heps : Filter.Tendsto eps Filter.atTop (nhds 0)) (hf : UniformContinuousOn f K) (hfbdd : BddAbove (f '' K)) :
Filter.Tendsto (fun (n : ) => (T n).sup' f) Filter.atTop (nhds (sSup (f '' K)))

Finite internal-net suprema converge to the supremum on K. The BddAbove premise is explicit because uniform continuity alone does not bound a function on an arbitrary set.

theorem HighDimProb.tendsto_finset_sup'_of_uniformContinuousOn_of_isInternalEpsilonNet {alpha : Type u_1} [PseudoMetricSpace alpha] {K : Set alpha} {T : Finset alpha} {eps : } {f : alpha} (hK : K.Nonempty) (hT : ∀ (n : ), (T n).Nonempty) (hnet : ∀ (n : ), IsInternalEpsilonNet K (↑(T n)) (eps n)) (heps_pos : ∀ (n : ), 0 < eps n) (heps : Filter.Tendsto eps Filter.atTop (nhds 0)) (hf : UniformContinuousOn f K) :
Filter.Tendsto (fun (n : ) => (T n).sup' f) Filter.atTop (nhds (sSup (f '' K)))

The bounded-above premise in the core theorem follows from uniform continuity and the same shrinking positive internal-net family.

theorem HighDimProb.tendsto_finset_sup'_abs_sub_of_uniformContinuousOn_of_isInternalEpsilonNet {alpha : Type u_1} [PseudoMetricSpace alpha] {K : Set alpha} {T : Finset alpha} {eps : } {f : alpha} {t0 : alpha} (ht0 : t0 K) (hT : ∀ (n : ), (T n).Nonempty) (hnet : ∀ (n : ), IsInternalEpsilonNet K (↑(T n)) (eps n)) (heps_pos : ∀ (n : ), 0 < eps n) (heps : Filter.Tendsto eps Filter.atTop (nhds 0)) (hf : UniformContinuousOn f K) :
Filter.Tendsto (fun (n : ) => (T n).sup' fun (t : alpha) => |f t - f t0|) Filter.atTop (nhds (sSup ((fun (t : alpha) => |f t - f t0|) '' K)))

Apply the supremum convergence theorem to the increment modulus rooted at t0 ∈ K.