A proximal-newton method for monotone inclusions in Hilbert spaces with complexity O(1/k 2 ).

Size: px
Start display at page:

Download "A proximal-newton method for monotone inclusions in Hilbert spaces with complexity O(1/k 2 )."

Transcription

1 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 A proximal-newton method for monotone inclusions in Hilbert spaces with complexity O(1/k 2 ). Hedy ATTOUCH Université Montpellier 2 ACSIOM, I3M UMR CNRS 5149 joint work with M. Marques Alves, and B.F. Svaiter Convegno Italo-Francese, Atelier Italo-Français Optimisation et processus dynamiques en apprentissage et problèmes inverses Fondazione Mediaterraneo, Sestri Levante (GE), Italia Sept. 8-12, 2014

2 1A. General presentation: dynamical approach H real Hilbert space; x 2 = x, x ; A : H H maximal monotone operator. Fast methods for solving: find x H such that 0 Ax. (1) (I + λa) 1 : H H resolvent of index λ > 0 of A. Fixed point formulation of (1): x (I + λa) 1 x = 0. Dynamical system: control variable: t λ(t). ẋ(t) + x(t) (I + λ(t)a) 1 x(t) = 0. (2) H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

3 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 1B. General presentation: dynamical approach Differential-algebraic system: ẋ(t) + x(t) (I + λ(t)a) 1 x(t) = 0, (LSP) λ(t) (I + λ(t)a) 1 x(t) x(t) = θ. (3) (LSP): Large Step Proximal method. (x( ), λ( )) variables, θ positive parameter. Closed-loop control: λ is taken inversely proportional to the speed. Asymptotic equilibration (t + ): ẋ(t) 0 (I + λ(t)a) 1 x(t) x(t) 0 λ(t) +.

4 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 1C. General presentation: dynamical approach Cauchy problem, main results: ẋ(t) + x(t) (I + λ(t)a) 1 x(t) = 0, (LSP) λ(t) (I + λ(t)a) 1 x(t) x(t) = θ, x(0) = x 0 H. 1 Existence and uniqueness of (x, λ) global solution of (LSP). 2 A 1 (0) : λ(t) +, w lim t + x(t) = x A 1 (0). 3 A = f, f : H R {+ } convex, lower semicontinuous, proper f (x(t)) inf H f = O( 1 t 2 ).

5 1D. General presentation: Large step proximal method A large-step proximal method for convex optimization: (0) x 0 dom(f ), σ [0, 1[, θ > 0 given, set k = 1; (1) choose λ k > 0, and find x k, v k H, ε k 0 such that v k εk f (x k ), (4) λ k v k + x k x k λ k ε k σ 2 x k x k 1 2, (5) λ k x k x k 1 θ or v k = 0; (6) (2) if v k = 0 then STOP and output x k ; otherwise let k k + 1 and go to step 1. end. Main result: f (x k ) inf H f C k 2. H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

6 Contents 1 General presentation. 2 Algebraic relationship linking λ and x. 3 Existence and uniqueness for the Cauchy problem. 4 Asymptotic behavior. 5 Link with the regularized Newton system. 6 The convex subdifferential case. 7 A large-step proximal method for convex optimization. 8 O( 1 ɛ ) proximal Newton method for convex optimization. 9 Perspective, open questions. 10 Appendix. Some examples H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

7 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 2A. Algebraic relationship linking λ and x λ (I + λa) 1 x x = θ. (7) Set ϕ : [0, [ H R +, ϕ(λ, x) := λ x (I + λa) 1 x. (8) Some classical results on resolvents, λ > 0, µ > 0, x H (1) Jλ A = (I + λa) 1 : H H nonexpansive, (2) Jλ Ax = ( µ JA µ λ x + ( 1 µ ) λ J A λ x ) ; (3) Jλ Ax JA µ x λ µ A λ x ; (4) lim λ 0 Jλ Ax = proj x; dom(a) (5) lim λ + Jλ Ax = proj A 1 (0)x, if A 1 (0).

8 2B. Algebraic relationship linking λ and x Properties of ϕ(λ, x) := λ x (I + λa) 1 x For any x 1, x 2 H and λ > 0, ϕ(λ, x 1 ) ϕ(λ, x 2 ) λ x 2 x 1. For any x H and 0 < λ 1 λ 2, λ 2 λ 1 ϕ(λ 1, x) ϕ(λ 2, x) and ϕ(λ, x) = 0 if and only if 0 A(x). ( λ2 λ 1 ) 2 ϕ(λ 1, x) (9) For any x / A 1 (0), λ [0, [ ϕ(λ, x) R + is continuous, strictly increasing, ϕ(0, x) = 0, and lim λ + ϕ(λ, x) = +. H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

9 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 2C. Algebraic relationship linking λ and x λ [0, [ ϕ(λ, x) R + continuous, strict. increasing, 0 +. Definition Λ θ : H \ A 1 (0) ]0, [, Λ θ (x) := ϕ(, x) 1 (θ). (10) λ ϕ(λ, x) θ 0 Λ θ (x) λ Example: A = rot(0; π 2 )

10 3A. Existence result for the Cauchy problem ẋ(t) + x(t) (I + λ(t)a) 1 x(t) = 0, λ(t) (I + λ(t)a) 1 x(t) x(t) = θ, x(0) = x 0 H \ A 1 (0). (11) ẋ(t) + x(t) (I + Λ θ (x(t))a) 1 x(t) = 0; x(0) = x 0. ẋ(t) = F (x(t)); x(0) = x 0. H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

11 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 3B. Existence result for the Cauchy problem F : Ω = H \ A 1 (0) H F (x) := J A Λ θ (x) x x. Continuity properties of F F : locally Lipschitz continuous. x H \ A 1 (0) 1 Λ θ (x): Lipschitz continuous with constant θ. Cauchy problem: ẋ(t) = F (x(t)); x(0) = x 0. Cauchy-Lipschitz theorem: local existence, and uniqueness. Global existence: estimate 0 λ( ) λ( ).

12 3C. Existence result for the Cauchy problem (LSP) ẋ(t) + x(t) (I + λ(t)a) 1 x(t) = 0, λ(t) (I + λ(t)a) 1 x(t) x(t) = θ, x(0) = x 0 H \ A 1 (0). Theorem 1 (existence and uniqueness) There exists a unique solution (x, λ) : [0, + [ H R ++ of (LSP); x( ) is C 1, and λ( ) is locally Lipschitz continuous. Moreover, (i) λ( ) is non-decreasing; 0 λ( ) λ( ); (ii) t Jλ(t) A x(t) x(t) is non-increasing. H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

13 4. Asymptotic behavior Theorem 2 (convergence) Suppose that A 1 (0). Let (x, λ) : [0, + [ H R ++ be the unique solution of the Cauchy problem (LSP) with x 0 H \ A 1 (0). Then, (i) ẋ(t) = x(t) J A λ(t) x(t) d 0/ 2t; hence lim t + ẋ(t) = 0; (ii) λ(t) θ d 0 2t; hence lim t + λ(t) = + ; (iii) w lim t + x(t) = x exists, for some x A 1 (0), where d 0 is the distance from x 0 to A 1 (0). Weak convergence: Opial s lemma, Fejer monotone property. Strong convergence: A strongly monotone; A = f, f inf-compact. H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

14 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 5A. Link with the regularized Newton system x 0 / A 1 (0), (x, λ) : [0, + [ H R ++ solution of (LSP). y(t) := (I + λ(t)a) 1 x(t), v(t) := 1 (x(t) y(t)). (12) λ(t) Claim: y( ) is solution of a regularized Newton system. Time derivation of λ(t)v(t) + y(t) x(t) = 0, and (LSP) gives v(t) Ay(t); ẏ(t) + λ(t) v(t) + (λ(t) + λ(t))v(t) = 0. (13)

15 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 5B. Link with the regularized Newton system Time rescaling Since 1 τ(t) = t 0 λ(u) + λ(u) λ(u) λ(u) + λ(u) du = t + ln(λ(t)/λ(0)). (14) λ(u) 2, t τ(t) 2t. Set y(t) = ỹ(τ(t)), v(t) = ṽ(τ(t)). { ṽ Aỹ; 1 λ τ 1 d dτ ỹ + d dτ ṽ + ṽ = 0. (15) Regularized Newton system [AS, SICON 2011]. Levenberg-Marquardt regularization parameter 1 λ τ 1 0 as τ +.

16 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 6A. The subdifferential case Suppose arg min f ; d 0 := inf{ x 0 z : z arg min f } = x 0 z. Theorem 3 (rate of convergence) Suppose that f (x(0)) < +. Then, (i) t f (x(t)) and t f (y(t)) are non-increasing; (ii) Set κ = θ/d0 3. For any t 0 f (x(t)) f ( z) [ f (x 0 ) f ( z) ] 1 + tκ 2 = O( 1 f (x 0 ) f ( z) t 2 ) κ f (x 0 ) f ( z) (16)

17 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 6B. The subdifferential case Differential inequality: d dt β cβ3/2, β(t) := f (x(t)) f ( z). Step 1: d f (x(t)) f (y(t)) f (x(t)). dt Integrate ẋ + x = y and apply Jensen s inequality (convexity of f ) f (x(t + h)) t+h f (y( )) non-increasing t [ e h f (x(t)) + (1 e h )f (y(u)) ] e u f (x(t + h)) e h f (x(t)) + (1 e h )f (y(t)) f (x(t + h)) f (x(t)) h 1 e h (f (y(t)) f (x(t)). h e t+h e t du.

18 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 6C. The subdifferential case Step 2: κ(f (x) f ( z)) 3/2 f (y) f (x) 1 + (3κ/2)(f (x 0 ) f ( z)) 1/2. Since v(t) f (y(t)), λ(t)v(t) = x(t) y(t), and λ(t) 2 v(t) = θ f (x(t)) f (y(t)) + x(t) y(t), v(t) f (y(t)) + λ(t) v(t) 2 Since v(t) f (y(t)), for any t 0 = f (y(t)) + θ v(t) 3/2. f (y(t)) f ( z) y(t) z, v(t) y(t) z v(t) x(t) z v(t) d 0 v(t) where we have used y(t) = J A λ(t) (x(t)), z = JA λ(t) ( z), JA λ(t) nonexpansive, and t x(t) z non-increasing. Combining the above inequalities f (x(t)) f (y(t)) + (f (y(t) f ( z)) 3/2 θ/d 3 0.

19 6D. The subdifferential case f (x(t)) f ( z) f (y(t)) f ( z) + (f (y(t) f ( z)) 3/2 θ/d 3 0. (17) Convexity of r r 3/2 : If a, b, c 0 and a b + cb 3/2 then b a ca 3/2 1 + (3c/2)a 1/2 Hence f (y) f (x) κ(f (x) f ( z)) 3/2, (18) 1 + (3κ/2)(f (x) f ( z)) 1/2 Since f (x( )) is non-increasing κ(f (x) f ( z)) 3/2 f (y) f (x). (19) 1 + (3κ/2)(f (x 0 ) f ( z)) 1/2 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

20 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 6E. The subdifferential case: y( ) versus x( ) y(t) = (I + λ(t) f ) 1 x(t) = J f λ(t) x(t) = prox λ(t)f x(t). (1) y( ): solution of a regularized Newton system. (2) x(t) J f λ(t) x(t) = ẋ(t) d 0/ 2t 0. w lim y(t) = w t + lim t + (3) f (x(t)) f λ(t) (x(t)) f (Jλ(t) f x(t)). Hence Hence x(t) arg min f. 0 f (y(t)) inf f f (x(t)) inf f = O( 1 H H t 2 ). (4) y( ) dom f : more (space) regularity than x(t). (5) f (y(t)) inf H f even if arg min f =.

21 7A. A large-step prox. method for convex optimization Algorithm 1: (0) x 0 dom(f ), σ [0, 1[, θ > 0 given, set k = 1; (1) choose λ k > 0, and find x k, v k H, ε k 0 such that v k εk f (x k ), (20) λ k v k + x k x k λ k ε k σ 2 x k x k 1 2, (21) λ k x k x k 1 θ or v k = 0; (22) (2) if v k = 0 STOP, output x k ; otherwise k k + 1, go to step 1. end Relative error for HPE, Solodov and Svaiter, SVVA, JCA (1999). Large-step condition, Montero and Svaiter, SIOPT (2010, 2012). H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

22 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 7B. A large-step prox. method for convex optimization Theorem 4 (complexity, ε k = 0) D 0 = sup{ x y max{f (x), f (y)} f (x 0 )} < +, κ 0 := θ(1 σ) f (x 0 ) f ( x) (i) f (x k ) f ( x) [ ] 2 = O(1/k 2 ). κ 0 f (x0 ) f ( x) 1 + k 2 + 3κ 0 f (x0 ) f ( x) (ii) for each k 2 even, there exists j {k/2 + 1,..., k} such that 4 v j 3 θ(1 σ) k 2+k f (x 0 ) f ( x) κ 0 f (x0 ) f ( x) 2 + 3κ 0 f (x0 ) f ( x) 2 2/3 D 3 0. = O(1/k 2 ).

23 8A. O( 1 ɛ ) prox. Newton method for convex optim. (P) minimize f (x) s.t. x H. AS1) f : H R convex, twice continuously differentiable; AS2) 2 f (x) 2 f (y) L x y x, y H; AS3) D 0 = sup{ y x max{f (x), f (y)} f (x 0 )} <. Algorithm 2 (0) x 0 H, 0 < σ l < σ u < 1 b given, set k = 1; (1) if f (x k 1 ) = 0 then stop. Otherwise, compute λ k > 0 s.t. 2σ l L λ k (I + λ k 2 f (x k 1 )) 1 λ k f (x k 1 ) 2σ u L ; (23) (2) set x k = x k 1 (I + λ k 2 f (x k 1 )) 1 λ k f (x k 1 ); (3) set k k + 1 and go to step 1. end. H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

24 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 8B. O( 1 ɛ ) prox. Newton method for convex optim. Proximal method y = (I + λ f ) 1 x λ f (y) + y x = 0. Perform a single Newton iteration from the current iterate x λ f (x) + (I + λ 2 f (x))(y x) = 0. i.e., y = x (I + λ 2 f (x)) 1 λ f (x). Hence the Proximal-Newton method x k = x k 1 (I + λ k 2 f (x k 1 )) 1 λ k f (x k 1 ).

25 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 8C. O( 1 ɛ ) prox. Newton method for convex optim. Theorem 5 (complexity) Suppose AS1), AS2), AS3) hold, {x k } generated by Algorithm 2. Let x be a solution of (P). For any given tolerance ε > 0 define κ 0 = 2σ l (1 σ u ) LD0 3, K = 2 + 3κ 0 f (x0 ) f ( x), κ 0 ε ) 2/3 2L (2 1/6 + 3κ 0 f (x0 ) f ( x) J = [2σ l (1 σ u )] 1/6 κ 1/3. 0 ε Then, the following statements hold: (a) for any k K, f (x k ) f ( x) ε; (b) there exists j 2 J such that f (x j ) ε.

26 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 8D. O( 1 ɛ ) prox. Newton method for convex optim. Solving w.r. λ > 0 2σ l L λ (I + λ 2 f (x k 1 )) 1 λ f (x k 1 ) 2σ u L ; The solution set is a closed interval [λ l, λ u ] s.t. λu λ l Binary search in ln λ may be used to find λ k. σu σ l. Φ(λ) := λ (I + λ 2 f (x)) 1 λ f (x), Ψ(µ) := Φ( 1 λ ). 2σ l L Ψ(µ) 2σ u L ; µ Ψ 2 (µ), and µ ln(ψ(µ)) are convex.

27 9A. Perspective, large step proximal methods Main result (LSP) is a large step proximal method with O( 1 k 2 ) convergence. Related results Compare with Nesterov steepest descent. Rockafellar, SICON 1976: A is α-strongly monotone, A 1 (0) = x, x k+1 x αλ k x k x. Superlinear convergence when λ k +. Att-Redont-Svaiter, JOTA 2013, regularized Newton method, Levenberg-Marquardt parameter µ k = 1 λ k = f (x k ) 1 3, yield convergence of (x k ) at the order 4 3. Other algebraic relation: λ (I + λa) 1 x x γ = θ. (LSP): γ = 1. H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

28 9B. Perspective, open questions Possible extensions Extension of f (x(t)) inf f = O( 1 t 2 ) and f (x k ) inf f = O( 1 k 2 ) to the non potential case: convex-concave saddle value problems? general maximal monotone operators (via Fitzpatrick function)? Combine (LSP) with splitting methods: forward-backward method, alternating proximal minimization? Implementation of the method and applications. Fast converging methods and second-order methods: in time (Nesterov, FISTA...), or space (Newton-like methods: LSP). Compare, combine them? (LSP) is linked to a regularized Newton method which still works in the nonconvex nonsmooth setting. Convergence in KL setting? H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

29 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 Appendix: Isotropic linear monotone operator α > 0 positive constant, A = αi, i.e., for every x H, Ax = αx. (λa + I ) 1 1 x = 1 + λα x (24) x (λa + I ) 1 x = λα x. 1 + λα (25) Given x 0 0, (LSP) can be written ẋ(t) + αλ(t) 1+αλ(t) x(t) = 0, λ(t) > 0, αλ(t) 2 1+αλ(t) x(t) = θ, x(0) = x 0 H \ A 1 (0). (26) Let us first integrate the above linear differential equation.

30 Appendix: Isotropic linear monotone operator We have x(t) = e (t) x 0 with (t) := t 0 αλ(τ) 1+αλ(τ) dτ. Hence αλ(t) αλ(t) e (t) = θ x 0. (27) First, check this equation at time t = 0. Equivalently αλ(0) αλ(0) = θ x 0. (28) This equation defines uniquely λ(0) > 0, because ξ αξ2 1+αξ is strictly increasing from [0, + [ onto [0, + [. Thus, the only thing we have to prove is the existence of a positive function t λ(t) such that h(t) := αλ(t)2 1 + αλ(t) e (t) is constant on [0, + [. (29) Writing that the derivative h is identically zero on [0, + [, we obtain H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40

31 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 Appendix: Isotropic linear monotone operator After integration, we obtain λ (t)(αλ(t) + 2) αλ(t) 2 = 0. (30) α ln λ(t) 2 2 = αt + α ln λ(0) λ(t) λ(0). (31) Let us introduce the function g : ]0, + [ R g(ξ) = α ln ξ 2 ξ. (32) As t increases from 0 to +, g(t) is strictly increasing from to +. Thus, for each t > 0, (31) has a unique solution λ(t) > 0. Moreover, t λ(t) is increasing, continuously differentiable, and lim t λ(t) = +. Returning to (31), we obtain that λ(t) e t.

32 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 Appendix: Antisymmetric linear monotone operator H = R 2, A = rot(0, π 2, A = A, A(ξ, η) = ( η, ξ). (λa + I ) 1 x = 1 ( ) 1 + λ 2 ξ + λη, η λξ (33) x (λa + I ) 1 x = λ ( ) 1 + λ 2 λξ η, λη + ξ. (34) ξ(t) + η(t) + λ(t) λ(t) 2 λ(t) ( ) 1 + λ(t) 2 λ(t)ξ(t) η(t) = 0, λ(t) > 0, (35) ) λ(t) ( 1 + λ(t) 2 λ(t)η(t) + ξ(t) = 0, λ(t) > 0, (36) ξ(t) 2 + η(t) 2 = θ, (37)

33 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 Appendix: Antisymmetric linear monotone operator Set u(t) = ξ(t) 2 + η(t) 2. After multiplying (35) by ξ(t), and multiplying (36) by η(t), then adding the results, we obtain Set We have Equation (37) becomes u (t) + 2λ(t)2 u(t) = λ(t) (t) := t 0 2λ(τ) 2 dτ. (38) λ(τ) u(t) = e (t) u(0). (39) λ(t) 2 e 1 (t) 2 = θ + λ(t) 2 x 0. (40)

34 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 Appendix: Antisymmetric linear monotone operator First, check this equation at time t = 0. Equivalently λ(0) 2 = θ 1 + λ(0) 2 x 0. (41) This equation defines uniquely λ(0) > 0, because the function ρ ρ2 is strictly increasing from [0, + [ onto [0, + [. Thus, we 1+ρ 2 just need to prove the existence of a positive function t λ(t) s.t. h(t) := λ(t) 2 e 1 (t) 2 is constant on [0, + [. (42) + λ(t) 2 Writing that the derivative h is identically zero on [0, + [, we obtain that λ( ) must satisfy λ (t)(2λ(t) + λ(t) 3 ) λ(t) 3 = 0. (43)

35 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 Appendix: Antisymmetric linear monotone operator After integration of this first-order differential equation, with Cauchy data λ(0), we obtain λ(t) 2 λ(t) = t + λ(0) 2 λ(0). (44) Let us introduce the function g : ]0, + [ R g(ρ) = ρ 2 ρ. (45) As t increases from 0 to +, g(t) is strictly increasing from to +. Thus, for each t > 0, (44) has a unique solution λ(t) > 0. Moreover, the mapping t λ(t) is increasing, continuously differentiable, and lim t λ(t) = +. Returning to (44), we obtain that λ(t) t as t +.

36 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 References B. Abbas, H. Attouch, and B. F. Svaiter, Newton-like dynamics and forward-backward methods for structured monotone inclusions in Hilbert spaces, JOTA, DOI /s , (2013). H. Attouch, Viscosity solutions of minimization problems, SIAM J. Optim., 6 (1996), No. 3, pp H. Attouch, J. Bolte, and B. F. Svaiter, Convergence of descent methods for semi-algebraic and tame problems: proximal algorithms, forward-backward splitting, and regularized Gauss-Seidel methods, Math. Program., 137 (2013), No. 1, pp H. Attouch, P. Redont, and B. F. Svaiter, Global convergence of a closed-loop regularized Newton method for solving monotone inclusions in Hilbert spaces, JOTA, 157 (2013), pp

37 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 References H. Attouch and B. F. Svaiter, A continuous dynamical Newton-like approach to solving monotone inclusions, SIAM J. Control Optim., 49 (2011), pp H. Bauschke and P. Combettes, Convex analysis and monotone operator theory, CMS books in Mathematics, Springer, H. Brézis, Opérateurs maximaux monotones et semi-groupes de contractions dans les espaces de Hilbert, North-Holland/Elsevier, New-York, R.E. Bruck, Asymptotic convergence of nonlinear contraction semigroups in Hilbert spaces, J. Funct. Anal., 18 (1975), pp

38 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 References C. Gonzaga, E. W. Karas, Fine tuning Nesterov s steepest descent algorithm for differentiable convex programming, Math. Program., 138 (2013), pp A. Griewank, The modification of Newton s method for unconstrained optimization by bounding cubic terms, Technical Report NA/12, Department of Applied Mathematics and Theoretical Physics, University of Cambridge, B. Martinet, Régularisation d inéquations variationnelles par approximations successives, Rev. Française Informat. Recherche Opérationnelle, 4 (1970), (Ser. R-3), pp R. D. C. Monteiro and B. F. Svaiter, On the complexity of the hybrid proximal extragradient method for the iterates and the ergodic mean, SIAM J. Optim., 20 (2010), No. 6, pp

39 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 References R. D. C. Monteiro and B. F. Svaiter, Iteration-complexity of a Newton proximal extragradient method for monotone variational inequalities and inclusion problems, SIAM J. Optim., 22 (2012), No. 3, pp Y. Nesterov, Introductory lectures on convex optimization, A basic course, Kluwer, Boston (2004). Y. Nesterov and B. T. Polyak, Cubic regularization of Newton method and its global performance, Math. Program., 108 (2006), (1, Ser. A), pp Z. Opial, Weak convergence of the sequence of successive approximations for nonexpansive mappings. Bull. Amer. Math. Soc. 73 (1967), pp

40 H. ATTOUCH (Univ. Montpellier 2) Fast proximal-newton method Sept. 8-12, / 40 References R. T. Rockafellar, Augmented Lagrangians and applications of the proximal point algorithm in convex programming, Math. Oper. Res., 1 (1976), No. 2, pp R. T. Rockafellar, Monotone operators and the proximal point algorithm, SIAM J. Control Optim., 14 (1976), No. 5, pp M. V. Solodov and B. F. Svaiter, A hybrid approximate extragradient-proximal point algorithm using the enlargement of a maximal monotone operator. Set-Valued Anal., 7 (1999), No. 4, pp M. V. Solodov and B. F. Svaiter, A hybrid projection-proximal point algorithm. J. Convex Anal., 6 (1999), No. 1, pp M. Weiser, P. Deuflhard, and B. Erdmann, Affine conjugate adaptive Newton methods for nonlinear elastomechanics, Optim. Methods Softw., 22 (2007), No. 3, pp

Une méthode proximale pour les inclusions monotones dans les espaces de Hilbert, avec complexité O(1/k 2 ).

Une méthode proximale pour les inclusions monotones dans les espaces de Hilbert, avec complexité O(1/k 2 ). Une méthode proximale pour les inclusions monotones dans les espaces de Hilbert, avec complexité O(1/k 2 ). Hedy ATTOUCH Université Montpellier 2 ACSIOM, I3M UMR CNRS 5149 Travail en collaboration avec

More information

Convergence rate estimates for the gradient differential inclusion

Convergence rate estimates for the gradient differential inclusion Convergence rate estimates for the gradient differential inclusion Osman Güler November 23 Abstract Let f : H R { } be a proper, lower semi continuous, convex function in a Hilbert space H. The gradient

More information

Convergence rate of inexact proximal point methods with relative error criteria for convex optimization

Convergence rate of inexact proximal point methods with relative error criteria for convex optimization Convergence rate of inexact proximal point methods with relative error criteria for convex optimization Renato D. C. Monteiro B. F. Svaiter August, 010 Revised: December 1, 011) Abstract In this paper,

More information

An Accelerated Hybrid Proximal Extragradient Method for Convex Optimization and its Implications to Second-Order Methods

An Accelerated Hybrid Proximal Extragradient Method for Convex Optimization and its Implications to Second-Order Methods An Accelerated Hybrid Proximal Extragradient Method for Convex Optimization and its Implications to Second-Order Methods Renato D.C. Monteiro B. F. Svaiter May 10, 011 Revised: May 4, 01) Abstract This

More information

Iteration-Complexity of a Newton Proximal Extragradient Method for Monotone Variational Inequalities and Inclusion Problems

Iteration-Complexity of a Newton Proximal Extragradient Method for Monotone Variational Inequalities and Inclusion Problems Iteration-Complexity of a Newton Proximal Extragradient Method for Monotone Variational Inequalities and Inclusion Problems Renato D.C. Monteiro B. F. Svaiter April 14, 2011 (Revised: December 15, 2011)

More information

c 2013 Society for Industrial and Applied Mathematics

c 2013 Society for Industrial and Applied Mathematics SIAM J. OPTIM. Vol. 3, No., pp. 109 115 c 013 Society for Industrial and Applied Mathematics AN ACCELERATED HYBRID PROXIMAL EXTRAGRADIENT METHOD FOR CONVEX OPTIMIZATION AND ITS IMPLICATIONS TO SECOND-ORDER

More information

arxiv: v3 [math.oc] 18 Apr 2012

arxiv: v3 [math.oc] 18 Apr 2012 A class of Fejér convergent algorithms, approximate resolvents and the Hybrid Proximal-Extragradient method B. F. Svaiter arxiv:1204.1353v3 [math.oc] 18 Apr 2012 Abstract A new framework for analyzing

More information

An Accelerated Hybrid Proximal Extragradient Method for Convex Optimization and its Implications to Second-Order Methods

An Accelerated Hybrid Proximal Extragradient Method for Convex Optimization and its Implications to Second-Order Methods An Accelerated Hybrid Proximal Extragradient Method for Convex Optimization and its Implications to Second-Order Methods Renato D.C. Monteiro B. F. Svaiter May 10, 011 Abstract This paper presents an accelerated

More information

A GENERALIZATION OF THE REGULARIZATION PROXIMAL POINT METHOD

A GENERALIZATION OF THE REGULARIZATION PROXIMAL POINT METHOD A GENERALIZATION OF THE REGULARIZATION PROXIMAL POINT METHOD OGANEDITSE A. BOIKANYO AND GHEORGHE MOROŞANU Abstract. This paper deals with the generalized regularization proximal point method which was

More information

A Dykstra-like algorithm for two monotone operators

A Dykstra-like algorithm for two monotone operators A Dykstra-like algorithm for two monotone operators Heinz H. Bauschke and Patrick L. Combettes Abstract Dykstra s algorithm employs the projectors onto two closed convex sets in a Hilbert space to construct

More information

A Lyusternik-Graves Theorem for the Proximal Point Method

A Lyusternik-Graves Theorem for the Proximal Point Method A Lyusternik-Graves Theorem for the Proximal Point Method Francisco J. Aragón Artacho 1 and Michaël Gaydu 2 Abstract We consider a generalized version of the proximal point algorithm for solving the perturbed

More information

Brøndsted-Rockafellar property of subdifferentials of prox-bounded functions. Marc Lassonde Université des Antilles et de la Guyane

Brøndsted-Rockafellar property of subdifferentials of prox-bounded functions. Marc Lassonde Université des Antilles et de la Guyane Conference ADGO 2013 October 16, 2013 Brøndsted-Rockafellar property of subdifferentials of prox-bounded functions Marc Lassonde Université des Antilles et de la Guyane Playa Blanca, Tongoy, Chile SUBDIFFERENTIAL

More information

On the complexity of the hybrid proximal extragradient method for the iterates and the ergodic mean

On the complexity of the hybrid proximal extragradient method for the iterates and the ergodic mean On the complexity of the hybrid proximal extragradient method for the iterates and the ergodic mean Renato D.C. Monteiro B. F. Svaiter March 17, 2009 Abstract In this paper we analyze the iteration-complexity

More information

Approaching monotone inclusion problems via second order dynamical systems with linear and anisotropic damping

Approaching monotone inclusion problems via second order dynamical systems with linear and anisotropic damping March 0, 206 3:4 WSPC Proceedings - 9in x 6in secondorderanisotropicdamping206030 page Approaching monotone inclusion problems via second order dynamical systems with linear and anisotropic damping Radu

More information

Algorithmes de minimisation alternée avec frottement: H. Attouch

Algorithmes de minimisation alternée avec frottement: H. Attouch Algorithmes de minimisation alternée avec frottement: Applications aux E.D.P., problèmes inverses et jeux dynamiques. H. Attouch Institut de Mathématiques et de Modélisation de Montpellier UMR CNRS 5149

More information

Variable Metric Forward-Backward Algorithm

Variable Metric Forward-Backward Algorithm Variable Metric Forward-Backward Algorithm 1/37 Variable Metric Forward-Backward Algorithm for minimizing the sum of a differentiable function and a convex function E. Chouzenoux in collaboration with

More information

ON A HYBRID PROXIMAL POINT ALGORITHM IN BANACH SPACES

ON A HYBRID PROXIMAL POINT ALGORITHM IN BANACH SPACES U.P.B. Sci. Bull., Series A, Vol. 80, Iss. 3, 2018 ISSN 1223-7027 ON A HYBRID PROXIMAL POINT ALGORITHM IN BANACH SPACES Vahid Dadashi 1 In this paper, we introduce a hybrid projection algorithm for a countable

More information

1 Introduction and preliminaries

1 Introduction and preliminaries Proximal Methods for a Class of Relaxed Nonlinear Variational Inclusions Abdellatif Moudafi Université des Antilles et de la Guyane, Grimaag B.P. 7209, 97275 Schoelcher, Martinique abdellatif.moudafi@martinique.univ-ag.fr

More information

Splitting Techniques in the Face of Huge Problem Sizes: Block-Coordinate and Block-Iterative Approaches

Splitting Techniques in the Face of Huge Problem Sizes: Block-Coordinate and Block-Iterative Approaches Splitting Techniques in the Face of Huge Problem Sizes: Block-Coordinate and Block-Iterative Approaches Patrick L. Combettes joint work with J.-C. Pesquet) Laboratoire Jacques-Louis Lions Faculté de Mathématiques

More information

I P IANO : I NERTIAL P ROXIMAL A LGORITHM FOR N ON -C ONVEX O PTIMIZATION

I P IANO : I NERTIAL P ROXIMAL A LGORITHM FOR N ON -C ONVEX O PTIMIZATION I P IANO : I NERTIAL P ROXIMAL A LGORITHM FOR N ON -C ONVEX O PTIMIZATION Peter Ochs University of Freiburg Germany 17.01.2017 joint work with: Thomas Brox and Thomas Pock c 2017 Peter Ochs ipiano c 1

More information

Strong Convergence Theorem by a Hybrid Extragradient-like Approximation Method for Variational Inequalities and Fixed Point Problems

Strong Convergence Theorem by a Hybrid Extragradient-like Approximation Method for Variational Inequalities and Fixed Point Problems Strong Convergence Theorem by a Hybrid Extragradient-like Approximation Method for Variational Inequalities and Fixed Point Problems Lu-Chuan Ceng 1, Nicolas Hadjisavvas 2 and Ngai-Ching Wong 3 Abstract.

More information

Steepest descent method on a Riemannian manifold: the convex case

Steepest descent method on a Riemannian manifold: the convex case Steepest descent method on a Riemannian manifold: the convex case Julien Munier Abstract. In this paper we are interested in the asymptotic behavior of the trajectories of the famous steepest descent evolution

More information

Sequential convex programming,: value function and convergence

Sequential convex programming,: value function and convergence Sequential convex programming,: value function and convergence Edouard Pauwels joint work with Jérôme Bolte Journées MODE Toulouse March 23 2016 1 / 16 Introduction Local search methods for finite dimensional

More information

Maximal Monotone Operators with a Unique Extension to the Bidual

Maximal Monotone Operators with a Unique Extension to the Bidual Journal of Convex Analysis Volume 16 (2009), No. 2, 409 421 Maximal Monotone Operators with a Unique Extension to the Bidual M. Marques Alves IMPA, Estrada Dona Castorina 110, 22460-320 Rio de Janeiro,

More information

1. Gradient method. gradient method, first-order methods. quadratic bounds on convex functions. analysis of gradient method

1. Gradient method. gradient method, first-order methods. quadratic bounds on convex functions. analysis of gradient method L. Vandenberghe EE236C (Spring 2016) 1. Gradient method gradient method, first-order methods quadratic bounds on convex functions analysis of gradient method 1-1 Approximate course outline First-order

More information

Iterative Convex Optimization Algorithms; Part One: Using the Baillon Haddad Theorem

Iterative Convex Optimization Algorithms; Part One: Using the Baillon Haddad Theorem Iterative Convex Optimization Algorithms; Part One: Using the Baillon Haddad Theorem Charles Byrne (Charles Byrne@uml.edu) http://faculty.uml.edu/cbyrne/cbyrne.html Department of Mathematical Sciences

More information

PROX-PENALIZATION AND SPLITTING METHODS FOR CONSTRAINED VARIATIONAL PROBLEMS

PROX-PENALIZATION AND SPLITTING METHODS FOR CONSTRAINED VARIATIONAL PROBLEMS PROX-PENALIZATION AND SPLITTING METHODS FOR CONSTRAINED VARIATIONAL PROBLEMS HÉDY ATTOUCH, MARC-OLIVIER CZARNECKI & JUAN PEYPOUQUET Abstract. This paper is concerned with the study of a class of prox-penalization

More information

Visco-penalization of the sum of two monotone operators

Visco-penalization of the sum of two monotone operators Visco-penalization of the sum of two monotone operators Patrick L. Combettes a and Sever A. Hirstoaga b a Laboratoire Jacques-Louis Lions, Faculté de Mathématiques, Université Pierre et Marie Curie Paris

More information

Maximal monotone operators are selfdual vector fields and vice-versa

Maximal monotone operators are selfdual vector fields and vice-versa Maximal monotone operators are selfdual vector fields and vice-versa Nassif Ghoussoub Department of Mathematics, University of British Columbia, Vancouver BC Canada V6T 1Z2 nassif@math.ubc.ca February

More information

An accelerated non-euclidean hybrid proximal extragradient-type algorithm for convex concave saddle-point problems

An accelerated non-euclidean hybrid proximal extragradient-type algorithm for convex concave saddle-point problems Optimization Methods and Software ISSN: 1055-6788 (Print) 1029-4937 (Online) Journal homepage: http://www.tandfonline.com/loi/goms20 An accelerated non-euclidean hybrid proximal extragradient-type algorithm

More information

Some Inexact Hybrid Proximal Augmented Lagrangian Algorithms

Some Inexact Hybrid Proximal Augmented Lagrangian Algorithms Some Inexact Hybrid Proximal Augmented Lagrangian Algorithms Carlos Humes Jr. a, Benar F. Svaiter b, Paulo J. S. Silva a, a Dept. of Computer Science, University of São Paulo, Brazil Email: {humes,rsilva}@ime.usp.br

More information

Viscosity approximation methods for the implicit midpoint rule of asymptotically nonexpansive mappings in Hilbert spaces

Viscosity approximation methods for the implicit midpoint rule of asymptotically nonexpansive mappings in Hilbert spaces Available online at www.tjnsa.com J. Nonlinear Sci. Appl. 9 016, 4478 4488 Research Article Viscosity approximation methods for the implicit midpoint rule of asymptotically nonexpansive mappings in Hilbert

More information

Second order forward-backward dynamical systems for monotone inclusion problems

Second order forward-backward dynamical systems for monotone inclusion problems Second order forward-backward dynamical systems for monotone inclusion problems Radu Ioan Boţ Ernö Robert Csetnek March 2, 26 Abstract. We begin by considering second order dynamical systems of the from

More information

M. Marques Alves Marina Geremia. November 30, 2017

M. Marques Alves Marina Geremia. November 30, 2017 Iteration complexity of an inexact Douglas-Rachford method and of a Douglas-Rachford-Tseng s F-B four-operator splitting method for solving monotone inclusions M. Marques Alves Marina Geremia November

More information

Iteration-complexity of a Rockafellar s proximal method of multipliers for convex programming based on second-order approximations

Iteration-complexity of a Rockafellar s proximal method of multipliers for convex programming based on second-order approximations Iteration-complexity of a Rockafellar s proximal method of multipliers for convex programming based on second-order approximations M. Marques Alves R.D.C. Monteiro Benar F. Svaiter February 1, 016 Abstract

More information

Second order forward-backward dynamical systems for monotone inclusion problems

Second order forward-backward dynamical systems for monotone inclusion problems Second order forward-backward dynamical systems for monotone inclusion problems Radu Ioan Boţ Ernö Robert Csetnek March 6, 25 Abstract. We begin by considering second order dynamical systems of the from

More information

An accelerated non-euclidean hybrid proximal extragradient-type algorithm for convex-concave saddle-point problems

An accelerated non-euclidean hybrid proximal extragradient-type algorithm for convex-concave saddle-point problems An accelerated non-euclidean hybrid proximal extragradient-type algorithm for convex-concave saddle-point problems O. Kolossoski R. D. C. Monteiro September 18, 2015 (Revised: September 28, 2016) Abstract

More information

On nonexpansive and accretive operators in Banach spaces

On nonexpansive and accretive operators in Banach spaces Available online at www.isr-publications.com/jnsa J. Nonlinear Sci. Appl., 10 (2017), 3437 3446 Research Article Journal Homepage: www.tjnsa.com - www.isr-publications.com/jnsa On nonexpansive and accretive

More information

On convergence rate of the Douglas-Rachford operator splitting method

On convergence rate of the Douglas-Rachford operator splitting method On convergence rate of the Douglas-Rachford operator splitting method Bingsheng He and Xiaoming Yuan 2 Abstract. This note provides a simple proof on a O(/k) convergence rate for the Douglas- Rachford

More information

Monotone variational inequalities, generalized equilibrium problems and fixed point methods

Monotone variational inequalities, generalized equilibrium problems and fixed point methods Wang Fixed Point Theory and Applications 2014, 2014:236 R E S E A R C H Open Access Monotone variational inequalities, generalized equilibrium problems and fixed point methods Shenghua Wang * * Correspondence:

More information

Journal of Convex Analysis (accepted for publication) A HYBRID PROJECTION PROXIMAL POINT ALGORITHM. M. V. Solodov and B. F.

Journal of Convex Analysis (accepted for publication) A HYBRID PROJECTION PROXIMAL POINT ALGORITHM. M. V. Solodov and B. F. Journal of Convex Analysis (accepted for publication) A HYBRID PROJECTION PROXIMAL POINT ALGORITHM M. V. Solodov and B. F. Svaiter January 27, 1997 (Revised August 24, 1998) ABSTRACT We propose a modification

More information

Coordinate Update Algorithm Short Course Proximal Operators and Algorithms

Coordinate Update Algorithm Short Course Proximal Operators and Algorithms Coordinate Update Algorithm Short Course Proximal Operators and Algorithms Instructor: Wotao Yin (UCLA Math) Summer 2016 1 / 36 Why proximal? Newton s method: for C 2 -smooth, unconstrained problems allow

More information

A WEAK-TO-STRONGCONVERGENCE PRINCIPLE FOR FEJÉR-MONOTONE METHODS IN HILBERT SPACES

A WEAK-TO-STRONGCONVERGENCE PRINCIPLE FOR FEJÉR-MONOTONE METHODS IN HILBERT SPACES MATHEMATICS OF OPERATIONS RESEARCH Vol. 26, No. 2, May 2001, pp. 248 264 Printed in U.S.A. A WEAK-TO-STRONGCONVERGENCE PRINCIPLE FOR FEJÉR-MONOTONE METHODS IN HILBERT SPACES HEINZ H. BAUSCHKE and PATRICK

More information

WEAK CONVERGENCE OF RESOLVENTS OF MAXIMAL MONOTONE OPERATORS AND MOSCO CONVERGENCE

WEAK CONVERGENCE OF RESOLVENTS OF MAXIMAL MONOTONE OPERATORS AND MOSCO CONVERGENCE Fixed Point Theory, Volume 6, No. 1, 2005, 59-69 http://www.math.ubbcluj.ro/ nodeacj/sfptcj.htm WEAK CONVERGENCE OF RESOLVENTS OF MAXIMAL MONOTONE OPERATORS AND MOSCO CONVERGENCE YASUNORI KIMURA Department

More information

Convergence of Fixed-Point Iterations

Convergence of Fixed-Point Iterations Convergence of Fixed-Point Iterations Instructor: Wotao Yin (UCLA Math) July 2016 1 / 30 Why study fixed-point iterations? Abstract many existing algorithms in optimization, numerical linear algebra, and

More information

An Inexact Spingarn s Partial Inverse Method with Applications to Operator Splitting and Composite Optimization

An Inexact Spingarn s Partial Inverse Method with Applications to Operator Splitting and Composite Optimization Noname manuscript No. (will be inserted by the editor) An Inexact Spingarn s Partial Inverse Method with Applications to Operator Splitting and Composite Optimization S. Costa Lima M. Marques Alves Received:

More information

Minimization vs. Null-Minimization: a Note about the Fitzpatrick Theory

Minimization vs. Null-Minimization: a Note about the Fitzpatrick Theory Minimization vs. Null-Minimization: a Note about the Fitzpatrick Theory Augusto Visintin Abstract. After a result of Fitzpatrick, for any maximal monotone operator α : V P(V ) there exists a function J

More information

arxiv: v2 [math.oc] 21 Nov 2017

arxiv: v2 [math.oc] 21 Nov 2017 Unifying abstract inexact convergence theorems and block coordinate variable metric ipiano arxiv:1602.07283v2 [math.oc] 21 Nov 2017 Peter Ochs Mathematical Optimization Group Saarland University Germany

More information

Complexity of the relaxed Peaceman-Rachford splitting method for the sum of two maximal strongly monotone operators

Complexity of the relaxed Peaceman-Rachford splitting method for the sum of two maximal strongly monotone operators Complexity of the relaxed Peaceman-Rachford splitting method for the sum of two maximal strongly monotone operators Renato D.C. Monteiro, Chee-Khian Sim November 3, 206 Abstract This paper considers the

More information

Asymptotic Convergence of the Steepest Descent Method for the Exponential Penalty in Linear Programming

Asymptotic Convergence of the Steepest Descent Method for the Exponential Penalty in Linear Programming Journal of Convex Analysis Volume 2 (1995), No.1/2, 145 152 Asymptotic Convergence of the Steepest Descent Method for the Exponential Penalty in Linear Programming R. Cominetti 1 Universidad de Chile,

More information

ITERATIVE SCHEMES FOR APPROXIMATING SOLUTIONS OF ACCRETIVE OPERATORS IN BANACH SPACES SHOJI KAMIMURA AND WATARU TAKAHASHI. Received December 14, 1999

ITERATIVE SCHEMES FOR APPROXIMATING SOLUTIONS OF ACCRETIVE OPERATORS IN BANACH SPACES SHOJI KAMIMURA AND WATARU TAKAHASHI. Received December 14, 1999 Scientiae Mathematicae Vol. 3, No. 1(2000), 107 115 107 ITERATIVE SCHEMES FOR APPROXIMATING SOLUTIONS OF ACCRETIVE OPERATORS IN BANACH SPACES SHOJI KAMIMURA AND WATARU TAKAHASHI Received December 14, 1999

More information

A convergence result for an Outer Approximation Scheme

A convergence result for an Outer Approximation Scheme A convergence result for an Outer Approximation Scheme R. S. Burachik Engenharia de Sistemas e Computação, COPPE-UFRJ, CP 68511, Rio de Janeiro, RJ, CEP 21941-972, Brazil regi@cos.ufrj.br J. O. Lopes Departamento

More information

On the iterate convergence of descent methods for convex optimization

On the iterate convergence of descent methods for convex optimization On the iterate convergence of descent methods for convex optimization Clovis C. Gonzaga March 1, 2014 Abstract We study the iterate convergence of strong descent algorithms applied to convex functions.

More information

Active sets, steepest descent, and smooth approximation of functions

Active sets, steepest descent, and smooth approximation of functions Active sets, steepest descent, and smooth approximation of functions Dmitriy Drusvyatskiy School of ORIE, Cornell University Joint work with Alex D. Ioffe (Technion), Martin Larsson (EPFL), and Adrian

More information

Splitting methods for decomposing separable convex programs

Splitting methods for decomposing separable convex programs Splitting methods for decomposing separable convex programs Philippe Mahey LIMOS - ISIMA - Université Blaise Pascal PGMO, ENSTA 2013 October 4, 2013 1 / 30 Plan 1 Max Monotone Operators Proximal techniques

More information

Variational inequalities for set-valued vector fields on Riemannian manifolds

Variational inequalities for set-valued vector fields on Riemannian manifolds Variational inequalities for set-valued vector fields on Riemannian manifolds Chong LI Department of Mathematics Zhejiang University Joint with Jen-Chih YAO Chong LI (Zhejiang University) VI on RM 1 /

More information

A user s guide to Lojasiewicz/KL inequalities

A user s guide to Lojasiewicz/KL inequalities Other A user s guide to Lojasiewicz/KL inequalities Toulouse School of Economics, Université Toulouse I SLRA, Grenoble, 2015 Motivations behind KL f : R n R smooth ẋ(t) = f (x(t)) or x k+1 = x k λ k f

More information

A splitting minimization method on geodesic spaces

A splitting minimization method on geodesic spaces A splitting minimization method on geodesic spaces J.X. Cruz Neto DM, Universidade Federal do Piauí, Teresina, PI 64049-500, BR B.P. Lima DM, Universidade Federal do Piauí, Teresina, PI 64049-500, BR P.A.

More information

FIXED POINTS IN THE FAMILY OF CONVEX REPRESENTATIONS OF A MAXIMAL MONOTONE OPERATOR

FIXED POINTS IN THE FAMILY OF CONVEX REPRESENTATIONS OF A MAXIMAL MONOTONE OPERATOR PROCEEDINGS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 00, Number 0, Pages 000 000 S 0002-9939(XX)0000-0 FIXED POINTS IN THE FAMILY OF CONVEX REPRESENTATIONS OF A MAXIMAL MONOTONE OPERATOR B. F. SVAITER

More information

Optimization for Machine Learning

Optimization for Machine Learning Optimization for Machine Learning (Problems; Algorithms - A) SUVRIT SRA Massachusetts Institute of Technology PKU Summer School on Data Science (July 2017) Course materials http://suvrit.de/teaching.html

More information

WEAK CONVERGENCE THEOREMS FOR EQUILIBRIUM PROBLEMS WITH NONLINEAR OPERATORS IN HILBERT SPACES

WEAK CONVERGENCE THEOREMS FOR EQUILIBRIUM PROBLEMS WITH NONLINEAR OPERATORS IN HILBERT SPACES Fixed Point Theory, 12(2011), No. 2, 309-320 http://www.math.ubbcluj.ro/ nodeacj/sfptcj.html WEAK CONVERGENCE THEOREMS FOR EQUILIBRIUM PROBLEMS WITH NONLINEAR OPERATORS IN HILBERT SPACES S. DHOMPONGSA,

More information

A General Iterative Method for Constrained Convex Minimization Problems in Hilbert Spaces

A General Iterative Method for Constrained Convex Minimization Problems in Hilbert Spaces A General Iterative Method for Constrained Convex Minimization Problems in Hilbert Spaces MING TIAN Civil Aviation University of China College of Science Tianjin 300300 CHINA tianming963@6.com MINMIN LI

More information

Master 2 MathBigData. 3 novembre CMAP - Ecole Polytechnique

Master 2 MathBigData. 3 novembre CMAP - Ecole Polytechnique Master 2 MathBigData S. Gaïffas 1 3 novembre 2014 1 CMAP - Ecole Polytechnique 1 Supervised learning recap Introduction Loss functions, linearity 2 Penalization Introduction Ridge Sparsity Lasso 3 Some

More information

BORIS MORDUKHOVICH Wayne State University Detroit, MI 48202, USA. Talk given at the SPCOM Adelaide, Australia, February 2015

BORIS MORDUKHOVICH Wayne State University Detroit, MI 48202, USA. Talk given at the SPCOM Adelaide, Australia, February 2015 CODERIVATIVE CHARACTERIZATIONS OF MAXIMAL MONOTONICITY BORIS MORDUKHOVICH Wayne State University Detroit, MI 48202, USA Talk given at the SPCOM 2015 Adelaide, Australia, February 2015 Based on joint papers

More information

ITERATIVE ALGORITHMS WITH ERRORS FOR ZEROS OF ACCRETIVE OPERATORS IN BANACH SPACES. Jong Soo Jung. 1. Introduction

ITERATIVE ALGORITHMS WITH ERRORS FOR ZEROS OF ACCRETIVE OPERATORS IN BANACH SPACES. Jong Soo Jung. 1. Introduction J. Appl. Math. & Computing Vol. 20(2006), No. 1-2, pp. 369-389 Website: http://jamc.net ITERATIVE ALGORITHMS WITH ERRORS FOR ZEROS OF ACCRETIVE OPERATORS IN BANACH SPACES Jong Soo Jung Abstract. The iterative

More information

Convergence Theorems of Approximate Proximal Point Algorithm for Zeroes of Maximal Monotone Operators in Hilbert Spaces 1

Convergence Theorems of Approximate Proximal Point Algorithm for Zeroes of Maximal Monotone Operators in Hilbert Spaces 1 Int. Journal of Math. Analysis, Vol. 1, 2007, no. 4, 175-186 Convergence Theorems of Approximate Proximal Point Algorithm for Zeroes of Maximal Monotone Operators in Hilbert Spaces 1 Haiyun Zhou Institute

More information

A derivative-free nonmonotone line search and its application to the spectral residual method

A derivative-free nonmonotone line search and its application to the spectral residual method IMA Journal of Numerical Analysis (2009) 29, 814 825 doi:10.1093/imanum/drn019 Advance Access publication on November 14, 2008 A derivative-free nonmonotone line search and its application to the spectral

More information

Iterative common solutions of fixed point and variational inequality problems

Iterative common solutions of fixed point and variational inequality problems Available online at www.tjnsa.com J. Nonlinear Sci. Appl. 9 (2016), 1882 1890 Research Article Iterative common solutions of fixed point and variational inequality problems Yunpeng Zhang a, Qing Yuan b,

More information

Closed-Loop Impulse Control of Oscillating Systems

Closed-Loop Impulse Control of Oscillating Systems Closed-Loop Impulse Control of Oscillating Systems A. N. Daryin and A. B. Kurzhanski Moscow State (Lomonosov) University Faculty of Computational Mathematics and Cybernetics Periodic Control Systems, 2007

More information

Self-dual Smooth Approximations of Convex Functions via the Proximal Average

Self-dual Smooth Approximations of Convex Functions via the Proximal Average Chapter Self-dual Smooth Approximations of Convex Functions via the Proximal Average Heinz H. Bauschke, Sarah M. Moffat, and Xianfu Wang Abstract The proximal average of two convex functions has proven

More information

An inertial forward-backward algorithm for the minimization of the sum of two nonconvex functions

An inertial forward-backward algorithm for the minimization of the sum of two nonconvex functions An inertial forward-backward algorithm for the minimization of the sum of two nonconvex functions Radu Ioan Boţ Ernö Robert Csetnek Szilárd Csaba László October, 1 Abstract. We propose a forward-backward

More information

Existence and Approximation of Fixed Points of. Bregman Nonexpansive Operators. Banach Spaces

Existence and Approximation of Fixed Points of. Bregman Nonexpansive Operators. Banach Spaces Existence and Approximation of Fixed Points of in Reflexive Banach Spaces Department of Mathematics The Technion Israel Institute of Technology Haifa 22.07.2010 Joint work with Prof. Simeon Reich General

More information

Error bounds for proximal point subproblems and associated inexact proximal point algorithms

Error bounds for proximal point subproblems and associated inexact proximal point algorithms Error bounds for proximal point subproblems and associated inexact proximal point algorithms M. V. Solodov B. F. Svaiter Instituto de Matemática Pura e Aplicada, Estrada Dona Castorina 110, Jardim Botânico,

More information

From error bounds to the complexity of first-order descent methods for convex functions

From error bounds to the complexity of first-order descent methods for convex functions From error bounds to the complexity of first-order descent methods for convex functions Nguyen Trong Phong-TSE Joint work with Jérôme Bolte, Juan Peypouquet, Bruce Suter. Toulouse, 23-25, March, 2016 Journées

More information

Convergence rates for an inertial algorithm of gradient type associated to a smooth nonconvex minimization

Convergence rates for an inertial algorithm of gradient type associated to a smooth nonconvex minimization Convergence rates for an inertial algorithm of gradient type associated to a smooth nonconvex minimization Szilárd Csaba László November, 08 Abstract. We investigate an inertial algorithm of gradient type

More information

Optimization methods

Optimization methods Lecture notes 3 February 8, 016 1 Introduction Optimization methods In these notes we provide an overview of a selection of optimization methods. We focus on methods which rely on first-order information,

More information

Hedy Attouch, Jérôme Bolte, Benar Svaiter. To cite this version: HAL Id: hal

Hedy Attouch, Jérôme Bolte, Benar Svaiter. To cite this version: HAL Id: hal Convergence of descent methods for semi-algebraic and tame problems: proximal algorithms, forward-backward splitting, and regularized Gauss-Seidel methods Hedy Attouch, Jérôme Bolte, Benar Svaiter To cite

More information

arxiv: v1 [math.oc] 20 Sep 2016

arxiv: v1 [math.oc] 20 Sep 2016 General Viscosity Implicit Midpoint Rule For Nonexpansive Mapping Shuja Haider Rizvi Department of Mathematics, Babu Banarasi Das University, Lucknow 68, India arxiv:169.616v1 [math.oc] Sep 16 Abstract:

More information

STRONG CONVERGENCE OF AN ITERATIVE METHOD FOR VARIATIONAL INEQUALITY PROBLEMS AND FIXED POINT PROBLEMS

STRONG CONVERGENCE OF AN ITERATIVE METHOD FOR VARIATIONAL INEQUALITY PROBLEMS AND FIXED POINT PROBLEMS ARCHIVUM MATHEMATICUM (BRNO) Tomus 45 (2009), 147 158 STRONG CONVERGENCE OF AN ITERATIVE METHOD FOR VARIATIONAL INEQUALITY PROBLEMS AND FIXED POINT PROBLEMS Xiaolong Qin 1, Shin Min Kang 1, Yongfu Su 2,

More information

A proximal-like algorithm for a class of nonconvex programming

A proximal-like algorithm for a class of nonconvex programming Pacific Journal of Optimization, vol. 4, pp. 319-333, 2008 A proximal-like algorithm for a class of nonconvex programming Jein-Shan Chen 1 Department of Mathematics National Taiwan Normal University Taipei,

More information

INERTIAL ACCELERATED ALGORITHMS FOR SOLVING SPLIT FEASIBILITY PROBLEMS. Yazheng Dang. Jie Sun. Honglei Xu

INERTIAL ACCELERATED ALGORITHMS FOR SOLVING SPLIT FEASIBILITY PROBLEMS. Yazheng Dang. Jie Sun. Honglei Xu Manuscript submitted to AIMS Journals Volume X, Number 0X, XX 200X doi:10.3934/xx.xx.xx.xx pp. X XX INERTIAL ACCELERATED ALGORITHMS FOR SOLVING SPLIT FEASIBILITY PROBLEMS Yazheng Dang School of Management

More information

Local strong convexity and local Lipschitz continuity of the gradient of convex functions

Local strong convexity and local Lipschitz continuity of the gradient of convex functions Local strong convexity and local Lipschitz continuity of the gradient of convex functions R. Goebel and R.T. Rockafellar May 23, 2007 Abstract. Given a pair of convex conjugate functions f and f, we investigate

More information

Cubic regularization of Newton s method for convex problems with constraints

Cubic regularization of Newton s method for convex problems with constraints CORE DISCUSSION PAPER 006/39 Cubic regularization of Newton s method for convex problems with constraints Yu. Nesterov March 31, 006 Abstract In this paper we derive efficiency estimates of the regularized

More information

A continuous gradient-like dynamical approach to Pareto-optimization in Hilbert spaces

A continuous gradient-like dynamical approach to Pareto-optimization in Hilbert spaces A continuous gradient-like dynamical approach to Pareto-optimization in Hilbert spaces Hedy Attouch, Xavier Goudou To cite this version: Hedy Attouch, Xavier Goudou. A continuous gradient-like dynamical

More information

Spectral gradient projection method for solving nonlinear monotone equations

Spectral gradient projection method for solving nonlinear monotone equations Journal of Computational and Applied Mathematics 196 (2006) 478 484 www.elsevier.com/locate/cam Spectral gradient projection method for solving nonlinear monotone equations Li Zhang, Weijun Zhou Department

More information

Thai Journal of Mathematics Volume 14 (2016) Number 1 : ISSN

Thai Journal of Mathematics Volume 14 (2016) Number 1 : ISSN Thai Journal of Mathematics Volume 14 (2016) Number 1 : 53 67 http://thaijmath.in.cmu.ac.th ISSN 1686-0209 A New General Iterative Methods for Solving the Equilibrium Problems, Variational Inequality Problems

More information

GENERAL NONCONVEX SPLIT VARIATIONAL INEQUALITY PROBLEMS. Jong Kyu Kim, Salahuddin, and Won Hee Lim

GENERAL NONCONVEX SPLIT VARIATIONAL INEQUALITY PROBLEMS. Jong Kyu Kim, Salahuddin, and Won Hee Lim Korean J. Math. 25 (2017), No. 4, pp. 469 481 https://doi.org/10.11568/kjm.2017.25.4.469 GENERAL NONCONVEX SPLIT VARIATIONAL INEQUALITY PROBLEMS Jong Kyu Kim, Salahuddin, and Won Hee Lim Abstract. In this

More information

Quasistatic Nonlinear Viscoelasticity and Gradient Flows

Quasistatic Nonlinear Viscoelasticity and Gradient Flows Quasistatic Nonlinear Viscoelasticity and Gradient Flows Yasemin Şengül University of Coimbra PIRE - OxMOS Workshop on Pattern Formation and Multiscale Phenomena in Materials University of Oxford 26-28

More information

On the convergence of a regularized Jacobi algorithm for convex optimization

On the convergence of a regularized Jacobi algorithm for convex optimization On the convergence of a regularized Jacobi algorithm for convex optimization Goran Banjac, Kostas Margellos, and Paul J. Goulart Abstract In this paper we consider the regularized version of the Jacobi

More information

An inexact subgradient algorithm for Equilibrium Problems

An inexact subgradient algorithm for Equilibrium Problems Volume 30, N. 1, pp. 91 107, 2011 Copyright 2011 SBMAC ISSN 0101-8205 www.scielo.br/cam An inexact subgradient algorithm for Equilibrium Problems PAULO SANTOS 1 and SUSANA SCHEIMBERG 2 1 DM, UFPI, Teresina,

More information

Regularization Inertial Proximal Point Algorithm for Convex Feasibility Problems in Banach Spaces

Regularization Inertial Proximal Point Algorithm for Convex Feasibility Problems in Banach Spaces Int. Journal of Math. Analysis, Vol. 3, 2009, no. 12, 549-561 Regularization Inertial Proximal Point Algorithm for Convex Feasibility Problems in Banach Spaces Nguyen Buong Vietnamse Academy of Science

More information

Higher-Order Methods

Higher-Order Methods Higher-Order Methods Stephen J. Wright 1 2 Computer Sciences Department, University of Wisconsin-Madison. PCMI, July 2016 Stephen Wright (UW-Madison) Higher-Order Methods PCMI, July 2016 1 / 25 Smooth

More information

Near Equality, Near Convexity, Sums of Maximally Monotone Operators, and Averages of Firmly Nonexpansive Mappings

Near Equality, Near Convexity, Sums of Maximally Monotone Operators, and Averages of Firmly Nonexpansive Mappings Mathematical Programming manuscript No. (will be inserted by the editor) Near Equality, Near Convexity, Sums of Maximally Monotone Operators, and Averages of Firmly Nonexpansive Mappings Heinz H. Bauschke

More information

Parameter Dependent Quasi-Linear Parabolic Equations

Parameter Dependent Quasi-Linear Parabolic Equations CADERNOS DE MATEMÁTICA 4, 39 33 October (23) ARTIGO NÚMERO SMA#79 Parameter Dependent Quasi-Linear Parabolic Equations Cláudia Buttarello Gentile Departamento de Matemática, Universidade Federal de São

More information

STRONG CONVERGENCE OF AN IMPLICIT ITERATION PROCESS FOR ASYMPTOTICALLY NONEXPANSIVE IN THE INTERMEDIATE SENSE MAPPINGS IN BANACH SPACES

STRONG CONVERGENCE OF AN IMPLICIT ITERATION PROCESS FOR ASYMPTOTICALLY NONEXPANSIVE IN THE INTERMEDIATE SENSE MAPPINGS IN BANACH SPACES Kragujevac Journal of Mathematics Volume 36 Number 2 (2012), Pages 237 249. STRONG CONVERGENCE OF AN IMPLICIT ITERATION PROCESS FOR ASYMPTOTICALLY NONEXPANSIVE IN THE INTERMEDIATE SENSE MAPPINGS IN BANACH

More information

Extensions of the CQ Algorithm for the Split Feasibility and Split Equality Problems

Extensions of the CQ Algorithm for the Split Feasibility and Split Equality Problems Extensions of the CQ Algorithm for the Split Feasibility Split Equality Problems Charles L. Byrne Abdellatif Moudafi September 2, 2013 Abstract The convex feasibility problem (CFP) is to find a member

More information

Steepest descent method on a Riemannian manifold: the convex case

Steepest descent method on a Riemannian manifold: the convex case Steepest descent method on a Riemannian manifold: the convex case Julien Munier To cite this version: Julien Munier. Steepest descent method on a Riemannian manifold: the convex case. 2006.

More information

LINEAR-CONVEX CONTROL AND DUALITY

LINEAR-CONVEX CONTROL AND DUALITY 1 LINEAR-CONVEX CONTROL AND DUALITY R.T. Rockafellar Department of Mathematics, University of Washington Seattle, WA 98195-4350, USA Email: rtr@math.washington.edu R. Goebel 3518 NE 42 St., Seattle, WA

More information

Duality and dynamics in Hamilton-Jacobi theory for fully convex problems of control

Duality and dynamics in Hamilton-Jacobi theory for fully convex problems of control Duality and dynamics in Hamilton-Jacobi theory for fully convex problems of control RTyrrell Rockafellar and Peter R Wolenski Abstract This paper describes some recent results in Hamilton- Jacobi theory

More information