§5, p. 408 — Algorithm MR is a special case of Algorithm GIN
ProvedGlobalInexactNewton.Backtracking.mr_isGINRunglobal-convergenceinexact-newtonp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a real normed space and . Every run of Algorithm MR with parameters , , , is a run of Algorithm GIN with the same , the same iterates , and the levels equal to the final values of the while-loop. That is, , and satisfies the inexact Newton condition (2.1) and the sufficient decrease condition (2.2) at level .
This is how the global convergence results for GIN (Theorems 3.4 and 3.5) transfer to Algorithm MR.
Formalization Note The final level of iteration is , the level visited after the loop body has run times. The statement needs no differentiability or finite dimension, so neither is assumed.
Preamble
import Mathlib import Definitions.Def_GlobalInexactNewton_Backtracking_Method open Filter Topology
Formal statement
namespace GlobalInexactNewton.Backtracking
/-- Eisenstat–Walker (1994), §5, before Theorem 5.2, p. 408: Algorithm MR is a special case of
Algorithm GIN. Every run of Algorithm MR is a run of Algorithm GIN with the same `t` and with the
final loop levels `η_k = trialLevel (ηbar k) (θ k) (m k)`. -/
theorem mr_isGINRun {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E]
(F : E → E) (ηmax t θmin θmax : ℝ) (x : ℕ → E) (ηbar : ℕ → ℝ) (σ : ℕ → ℝ → E)
(θ : ℕ → ℕ → ℝ) (m : ℕ → ℕ) (hrun : IsMRRun F ηmax t θmin θmax x ηbar σ θ m) :
IsGINRun F t x (fun k => trialLevel (ηbar k) (θ k) (m k)) := by sorry
end GlobalInexactNewton.Backtracking
Source
Eisenstat and Walker, Globally Convergent Inexact Newton Methods, SIAM J. Optim. 4(2) (1994), §5, before Theorem 5.2, p. 408
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.