(2.18) — Ψ(x_k) ≤ Ψ(x_{k−1}) − λ_k(1 − L_Ψλ_k)‖∇Ψ(x^md_k)‖² + (L_ΨΓ_k/2) Σ_{τ≤k} ((λ_τ−β_τ)²/(Γ_τα_τ))‖∇Ψ(x^md_τ)‖²
ProvedNonconvexAG.Smooth.eq_2_18accelerated-gradientnonconvex-optimizationp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be differentiable with -Lipschitz gradient, , and let be generated by Algorithm 1 from with step sizes , (), , ; as in (2.6). Then for every ,
Summing this one-step bound over gives (2.19).
Formalization Note The statement is the outer ends of the chain (2.18).
Preamble
import Mathlib import Definitions.Def_ConvexOptAlg_SmoothGD_Defs import Definitions.Def_NonconvexAG_Smooth_AGRun
Formal statement
namespace NonconvexAG.Smooth
/-- (2.18), p. 6 (outer ends of the chain). -/
theorem eq_2_18 {n : ℕ} (Ψ : E n → ℝ) (g : E n → E n) (LΨ : ℝ)
(hΨ : ConvexOptAlg.SmoothGD.IsBetaSmooth Ψ g LΨ) (hL : 0 < LΨ)
(α β lam : ℕ → ℝ) (hstep : AGStepsizes α β lam) (x0 : E n) :
∀ k, 1 ≤ k →
Ψ (xSeq g α β lam x0 k) ≤ Ψ (xSeq g α β lam x0 (k - 1)) - lam k * (1 - LΨ * lam k) * ‖g (xmdSeq g α β lam x0 k)‖ ^ 2
+ LΨ * Gamma α k / 2 * ∑ τ ∈ Finset.Icc 1 k,
(lam τ - β τ) ^ 2 / (Gamma α τ * α τ) * ‖g (xmdSeq g α β lam x0 τ)‖ ^ 2 := by sorry
end NonconvexAG.Smooth
Source
Ghadimi & Lan, Accelerated Gradient Methods for Nonconvex Nonlinear and Stochastic Programming, arXiv:1310.3787v1, p. 6, (2.18)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.