# Lecture-17 - 1 Lecture-17 Theorems 5.8 and 5.9&...

This preview shows pages 1–5. Sign up to view the full content.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 1 Lecture-17 Theorems 5.8 and 5.9 & Levenberg-Marquadet = k b ∑ = ∞ < ∇ K , , 2 2 1 || || k k k k f K , , 2 1 k k Since at the restarts 1 cos- = k q ∑ ≥ ∞ < ∇ 2 2 || || cos k k k f q We can use Theorem 5.7 to prove global convergence for algorithms, which are periodically started by setting If restarts occur at ∞ = 1 } { j j k If the restarts are done after every n iterations, the sequence is infinite || || lim = ∇ ∞ → j k j f Therefore a subsequence of gradients will approach to zero: || || inf lim = ∇ ∞ → k k f Convergence of Algorithms with restarts 2 Theorem 5.8 Suppose that the function is Lipschitz continuously differentiable, , and Algorithm 5.4 is implemented with a line search that satisfies strong Wolfe conditions, with 0< c 2 <1/2 . Then || || inf lim = ∇ ∞ → k k f g ≤ ∇ || || k f Proof Proof by contradiction: Assume that: large, ly sufficient all for || || ≥ ∇ g g k f k 1 , || || || || cos || || || || 2 1 K , k p f p f k k k k k = 2200 ∇ ≤ ≤ ∇ c q c ∑ ≥ ∞ < ∇ 2 2 || || cos k k k f q ∞ < ∇ ∑ ∞ = 2 4 1 || || || || k k k p f (D) || || inf lim = ∇ ∞ → k k f We have proved this (Lecture 16) Now 3 Proof ; 1 1 1 k FR k k k p f p + + + +-∇ ← b 1- +-∇ = k FR k k k p f p b 1 1 2 1 1 1 2 1 1 1 1 ) ( 2 ) ( ) ( ) (--------- + ∇- ∇ ∇ = + ∇- ∇- ∇ ∇ = +-∇ +-∇ = k T k FR k k T k FR k k T k k T k k T k FR k T k FR k k T k FR k k T k k T k k FR k k T k FR k k k T k p p p f f f p p p p f p p f f f p p p f p f p p b b b b b b b 2 1 2 2 1 1 2 1 || || 1 | |---- ∇- ≤ ∇- ≤ ∇ k k T k k T k f c c p f c p f k T k k T k FR k f f f f ∇ ∇ ∇ ∇ = + + + 1 1 1 b 2 1 2 2 2 2 2 2 1 2 2 1 1 1 2 2 2 2 1 2 2 1 2 2 2 2 1 2 1 2 2 || || ) ( || || 1 1 || || || || ) ( || || 1 2 || || || || ) ( || || 1 2 || || || || ) ( | | 2 || || || ||--------- + ∇ - + ≤ + ∇ ∇ ∇ ∇ ∇- + ∇ = + ∇- + ∇ ≤ + ∇ + ∇ ≤ k FR k k k k FR k k k T k k T k k k FR k k FR k k k FR k k T k FR k k k p f c c p p f f f f f c c f p f c c f p p f f p b b b b b b From (C) Proof k T k k T k p f c p f ∇- ≤ ∇ + 2 1 | | 1 , 1 1 2 || || 1 1 2 2 2 2 K , k c c f p f c k k T k = 2200-- ≤ ∇ ∇ ≤-- 1 1 2 1 | |--- ∇- ≤ ∇ k T k k T k p f c p f 1 1 2 || || 1 1 2 2 2 1 1 1 2 c c f p f c k k T k-- ≤ ∇ ∇ ≤----- 2 1 2 2 1 1 2 1 || || 1 | |---- ∇- ≤ ∇- ≤ ∇ k k T k k T k f c c p f c p f (A) (B) Combining (A) and (B) (C) Wolf’s condition Lemma 5.6 4 Proof 2 1 2 2 2 2 2 || || ) ( || || 1 1 || ||- + ∇ - + ≤ k FR k k k p f c c p b || || || || || || || || || || || || || || 1 || || 1 || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || || ) ( ) ( || || ) ( || || || || ] || || ) ( || || [ ) ( || || || || 1 2 4 3 2 2 2 4 2 4 2 1 2 4 3 2 2 4 2 4 3 2 1 4 2 3 2 2 2 4 2 4 1 4 1 4 2 1 3 4 1 4 2 3 2 2 2 2 1 2 2 1 3 2 2 3 2 2 2 2 1 2 1 3 2 2 3 2 ∑ =-...
View Full Document

## This note was uploaded on 06/12/2011 for the course COT 6505 taught by Professor Shah during the Spring '07 term at University of Central Florida.

### Page1 / 18

Lecture-17 - 1 Lecture-17 Theorems 5.8 and 5.9&...

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online