Documentation

HexRootsMathlib.Kantorovich

theorem NewtonKantorovich.ContractingWith.eq_of_fixedPoints {α : Type u_1} [MetricSpace α] {K : NNReal} {f : αα} {s : Set α} (hsf : Set.MapsTo f s s) (hf : ContractingWith K (Set.MapsTo.restrict f s s hsf)) {x y : α} (hxs : x s) (hys : y s) (hx : Function.IsFixedPt f x) (hy : Function.IsFixedPt f y) :
x = y

A map that contracts on a forward-invariant set has at most one fixed point in that set.

noncomputable def NewtonKantorovich.ContractingWith.fixedPoint' {α : Type u_1} [MetricSpace α] [CompleteSpace α] {K : NNReal} (f : αα) {s : Set α} (hsc : IsClosed s) (hs' : s.Nonempty) (hsf : Set.MapsTo f s s) (hf : ContractingWith K (Set.MapsTo.restrict f s s hsf)) :
α

Let s be a complete, nonempty, forward-invariant set of a self-map f. If f contracts on s, then fixedPoint' is the unique fixed point of the restriction of f to s.

Equations
Instances For
    theorem NewtonKantorovich.ContractingWith.fixedPoint'_mem {α : Type u_1} [MetricSpace α] [CompleteSpace α] {K : NNReal} {f : αα} {s : Set α} (hsc : IsClosed s) (hs' : s.Nonempty) (hsf : Set.MapsTo f s s) (hf : ContractingWith K (Set.MapsTo.restrict f s s hsf)) :
    fixedPoint' f hsc hs' hsf hf s

    The constructed fixed point lies in the invariant set.

    theorem NewtonKantorovich.ContractingWith.fixedPoint'_isFixedPt {α : Type u_1} [MetricSpace α] [CompleteSpace α] {K : NNReal} {f : αα} {s : Set α} (hsc : IsClosed s) (hs' : s.Nonempty) (hsf : Set.MapsTo f s s) (hf : ContractingWith K (Set.MapsTo.restrict f s s hsf)) :
    Function.IsFixedPt f (fixedPoint' f hsc hs' hsf hf)

    The constructed point is genuinely fixed by f.

    theorem NewtonKantorovich.ContractingWith.fixedPoint'_unique {α : Type u_1} [MetricSpace α] [CompleteSpace α] {K : NNReal} {f : αα} {s : Set α} (hsc : IsClosed s) (hs' : s.Nonempty) (hsf : Set.MapsTo f s s) (hf : ContractingWith K (Set.MapsTo.restrict f s s hsf)) {x : α} (hxs : x s) (hx : Function.IsFixedPt f x) :
    x = fixedPoint' f hsc hs' hsf hf

    Any fixed point of f in the invariant set is the constructed one.

    Membership in a closed ball, phrased with nnnorm.

    theorem NewtonKantorovich.contractingWith_of_nnnorm_fderiv_le {X : Type u_2} [NormedAddCommGroup X] [NormedSpace X] {T : XX} {DT : XX →L[] X} (hT : ∀ (x : X), HasFDerivAt T (DT x) x) {x₀ : X} {r : NNReal} (hTx₀ : Set.MapsTo T (Metric.closedBall x₀ r) (Metric.closedBall x₀ r)) {K : NNReal} (hzK : K < 1) (hDT : ∀ {x : X}, x Metric.closedBall x₀ rDT x‖₊ K) :

    Derivative criterion for ContractingWith.

    theorem NewtonKantorovich.image_bound {X : Type u_2} [NormedAddCommGroup X] [NormedSpace X] [CompleteSpace X] {T : XX} {DT : XX →L[] X} (hT : ∀ (x : X), HasFDerivAt T (DT x) x) (hDT : Continuous DT) {R : NNReal} {x₀ : X} {y z₁ z₂ : NNReal} (hy : T x₀ - x₀‖₊ y) (hz₁ : DT x₀‖₊ z₁) (hz₂ : xMetric.closedBall x₀ R, DT x - DT x₀‖₊ z₂ * x - x₀‖₊) {r : NNReal} (hrR : r R) {x : X} (hx : x Metric.closedBall x₀ r) :
    T x - x₀‖₊ y + z₁ * r + z₂ * r ^ 2 / 2

    Quantitative bound for the image of a point in a closed ball. This is the estimate that supplies the factor 1 / 2 in Newton--Kantorovich.

    theorem NewtonKantorovich.contraction_mapping {X : Type u_2} [NormedAddCommGroup X] [NormedSpace X] [CompleteSpace X] {T : XX} {DT : XX →L[] X} (hT : ∀ (x : X), HasFDerivAt T (DT x) x) (hDT : Continuous DT) {R : NNReal} {x₀ : X} {y z₁ z₂ : NNReal} (hy : T x₀ - x₀‖₊ y) (hz₁ : DT x₀‖₊ z₁) (hz₂ : xMetric.closedBall x₀ R, DT x - DT x₀‖₊ z₂ * x - x₀‖₊) {r : NNReal} (hrR : r R) (hyr : y + z₁ * r + z₂ * r ^ 2 / 2 r) (hzr : z₁ + z₂ * r < 1) :
    ∃! x : X, T x = x x - x₀‖₊ r

    Quantitative contraction mapping theorem on a closed ball.

    theorem NewtonKantorovich.newton_kantorovich {X : Type u_2} [NormedAddCommGroup X] [NormedSpace X] [CompleteSpace X] {Y : Type u_3} [NormedAddCommGroup Y] [NormedSpace Y] {F : XY} {DF : XX →L[] Y} (hF : ∀ (x : X), HasFDerivAt F (DF x) x) (hDF : Continuous DF) {A : Y →L[] X} (hA : Function.Injective A) {R : ENNReal} {x₀ : X} {y z₁ z₂ : NNReal} (hy : A (F x₀)‖₊ y) (hz₁ : 1 - A ∘SL DF x₀‖₊ z₁) (hz₂ : xMetric.closedEBall x₀ R, A ∘SL (DF x - DF x₀)‖₊ z₂ * x - x₀‖₊) {r : NNReal} (hrR : r R) (hyr : y + z₁ * r + z₂ * r ^ 2 / 2 r) (hzr : z₁ + z₂ * r < 1) :
    ∃! x : X, F x = 0 x - x₀‖₊ r

    Newton--Kantorovich theorem with an explicitly injective approximate inverse.

    theorem NewtonKantorovich.surjective {X : Type u_2} [NormedAddCommGroup X] [NormedSpace X] [CompleteSpace X] {Y : Type u_3} [NormedAddCommGroup Y] [NormedSpace Y] {DF : X →L[] Y} {A : Y →L[] X} {z₁ : NNReal} (hz₁ : 1 - A ∘SL DF‖₊ z₁) {r : NNReal} (hzr : z₁ + r < 1) :

    The approximate inverse is surjective when its composition error is small.

    theorem NewtonKantorovich.newton_kantorovich_fd {X : Type u_2} [NormedAddCommGroup X] [NormedSpace X] [CompleteSpace X] {Y : Type u_3} [NormedAddCommGroup Y] [NormedSpace Y] [FiniteDimensional X] [FiniteDimensional Y] (hXY : Module.finrank X = Module.finrank Y) {F : XY} {DF : XX →L[] Y} (hF : ∀ (x : X), HasFDerivAt F (DF x) x) (hDF : Continuous DF) {A : Y →L[] X} {R : ENNReal} {x₀ : X} {y z₁ z₂ : NNReal} (hy : A (F x₀)‖₊ y) (hz₁ : 1 - A ∘SL DF x₀‖₊ z₁) (hz₂ : xMetric.closedEBall x₀ R, A ∘SL (DF x - DF x₀)‖₊ z₂ * x - x₀‖₊) {r : NNReal} (hrR : r R) (hyr : y + z₁ * r + z₂ * r ^ 2 / 2 r) (hzr : z₁ + z₂ * r < 1) :
    ∃! x : X, F x = 0 x - x₀‖₊ r

    Finite-dimensional Newton--Kantorovich theorem; equal dimensions make the approximate inverse injective as a consequence of the smallness assumptions.