Documentation

LeanPool.ScottishBook155.RetractiveEnvelope

A retractive dual-evaluation envelope #

The dual-evaluation coordinate already gives the metric part of the relative envelope. Adjoining the old target as a max-product coordinate makes the retraction linear and explicit: it is first-coordinate projection.

@[reducible, inline]
abbrev ScottishBook155.RetractiveEnvelope (P : Type u) (N : Type v) [MetricSpace P] [NormedAddCommGroup N] [NormedSpace ℝ N] (j : N → P) :
Type (max v u v)

The old target together with the dual-evaluation relative coordinate.

Equations
Instances For
    noncomputable def ScottishBook155.retractiveEmbedding {P : Type u} {N : Type v} [MetricSpace P] [NormedAddCommGroup N] [NormedSpace ℝ N] (j : N → P) (R : P → N) (p : P) :

    Embed the attached metric space using a chosen metric retraction and the relative evaluation coordinate.

    Equations
    Instances For

      The old target embeds linearly in both coordinates.

      Equations
      Instances For

        The old target is a linear isometric subspace of the retractive envelope.

        Equations
        Instances For
          noncomputable def ScottishBook155.retractiveProjection {P : Type u} {N : Type v} [MetricSpace P] [NormedAddCommGroup N] [NormedSpace ℝ N] (j : N → P) :

          First-coordinate projection is the contractive linear retraction.

          Equations
          Instances For
            theorem ScottishBook155.retractiveEmbedding_target {P : Type u} {N : Type v} [MetricSpace P] [NormedAddCommGroup N] [NormedSpace ℝ N] (j : N → P) (R : P → N) (hR : ∀ (n : N), R (j n) = n) (n : N) :
            theorem ScottishBook155.retractiveProjection_embedding {P : Type u} {N : Type v} [MetricSpace P] [NormedAddCommGroup N] [NormedSpace ℝ N] (j : N → P) (R : P → N) (p : P) :
            theorem ScottishBook155.retractiveEmbedding_dist_le {P : Type u} {N : Type v} [MetricSpace P] [NormedAddCommGroup N] [NormedSpace ℝ N] (j : N → P) (R : P → N) (hR : ∀ (p q : P), dist (R p) (R q) ≤ dist p q) (p q : P) :

            A nonexpansive metric retraction and relative evaluation jointly give a nonexpansive embedding into the max-product envelope.

            theorem ScottishBook155.retractiveEmbedding_dist_eq {P : Type u} {N : Type v} [MetricSpace P] [NormedAddCommGroup N] [NormedSpace ℝ N] (j : N → P) (R : P → N) (hR : ∀ (p q : P), dist (R p) (R q) ≤ dist p q) {p q : P} (heval : dist (relativeEvaluation j p) (relativeEvaluation j q) = dist p q) :

            Every distance retained by relative evaluation remains exact in the retractive envelope.