Two Strong Convergence Theorems for a Proximal Method in Reflexive Banach Spaces

Size: px
Start display at page:

Download "Two Strong Convergence Theorems for a Proximal Method in Reflexive Banach Spaces"

Transcription

1 Two Strong Convergence Theorems for a Proxmal Method n Reflexve Banach Spaces Smeon Rech and Shoham Sabach Abstract. Two strong convergence theorems for a proxmal method for fndng common zeroes of maxmal monotone operators n reflexve Banach spaces are establshed. Both theorems take nto account possble computatonal errors. 1. Introducton In ths paper X denotes a real reflexve Banach space wth norm and X stands for the topologcal dual of X endowed wth the nduced norm. We denote the value of the functonal ξ X at x X by ξ, x. An operator A : X 2 X s sad to be monotone f for any x, y dom A, we have ξ Ax and η Ay = ξ η, x y 0. Recall that the set dom A = {x X : Ax } s called the effectve doman of such an operator A. A monotone operator A s sad to be maxmal f graph A, the graph of A, s not a proper subset of the graph of any other monotone operator. In ths paper f : X, + ] s always a proper, lower semcontnuous and convex functon, and f : X, + ] s the Fenchel conjugate of f. The set of nonnegatve ntegers wll be denoted by N. The problem of fndng an element x X such that 0 Ax s very mportant n Optmzaton Theory and related felds. For example, f A s the subdfferental f of f, then A s a maxmal monotone operator and the equaton 0 f x s equvalent to the problem of mnmzng f over X. One of the methods for solvng ths problem n Hlbert space s the well-known proxmal pont algorthm. Let H be a Hlbert space and let I denote the dentty operator on H. The proxmal pont algorthm generates, for any startng pont x 0 = x H, a sequence {x n } n N n H 2000 Mathematcs Subject Classfcaton. 47H05, 47J25. Key words and phrases. Banach space, Bregman projecton, Legendre functon, monotone operator, proxmal pont algorthm, resolvent, totally convex functon. 1

2 2 SIMEON REICH AND SHOHAM SABACH by the rule 1.1 x n+1 = I + λ n A 1 x n, n = 0, 1, 2,..., where {λ n } n N s a gven sequence of postve real numbers. equvalent to Note that 1.1 s 0 Ax n λ n x n+1 x n, n = 0, 1, 2,.... Ths algorthm was frst ntroduced by Martnet [28] and further developed by Rockafellar [38], who proves that the sequence generated by 1.1 converges weakly to an element of A 1 0 when A 1 0 s nonempty and lm nf n + λ n > 0. Furthermore, Rockafellar [38] asks f the sequence generated by 1.1 converges strongly. Ths queston was answered n the negatve by Güler [24], who presented an example of a subdfferental for whch the sequence generated by 1.1 converges weakly but not strongly; see [7] for a more recent and smpler example. Qute a few results regardng the proxmal pont algorthm and ts extensons can be found n the lterature. See, for example, [5, 6, 7, 10, 11, 15, 16, 19, 21, 22, 25, 26, 29, 30, 32, 33, 35, 39, 41, 43]. We menton, n partcular, the semnal papers [41, 21, 5, 6]. These papers ntroduce a new paradgm whch has snce led to many modfcatons. One such modfcaton has been proposed by Bauschke and Combettes [5] see also Solodov and Svater [41], who have modfed the proxmal pont algorthm n order to generate a strongly convergent sequence. They ntroduce, for example, the followng algorthm see [5, Corollary 6.1, p. 258] for a sngle operator and λ n = 1/2: x 0 H, 1.2 y n = R λna x n, C n = {z H : y n z x n z }, Q n = {z H : x 0 x n, z x n 0}, x n+1 = P Cn Q n x 0, n = 0, 1, 2,.... Here, for each x H and each nonempty, closed and convex subset C of H, the mappng P C s defned by x P C x = nf { x z : z C}. Ths mappng s called the metrc projecton of H onto C. The mappng R λa = I + λa 1 s the classcal resolvent of the maxmal monotone operator A. They prove that f A 1 0 s nonempty and lm nf n + λ n > 0, then the sequence generated by 1.2 converges strongly to P A 1 0. We and Zhou [42] generalze ths result to those Banach spaces X whch are both unformly convex and unformly smooth. They

3 STRONG CONVERGENCE THEOREMS 3 ntroduce the followng algorthm: x 0 X, 1.3 y n = J λn x n, C n = {z X : φ z, y n φ z, x n }, Q n = {z X : Jx 0 Jx n, z x n 0}, x n+1 = Q Cn Q n x 0, n = 0, 1, 2,..., where J s the normalzed dualty mappng of the space X, J λ x = J + λa 1 J and φ y, x = y 2 2 Jx, y + x 2. Here, for each nonempty, closed and convex subset C of X, Q C s a certan generalzaton of the metrc projecton P C n H. They prove that f A 1 0 s nonempty and lm nf n + λ n > 0, then the sequence generated by 1.3 converges strongly to Q A 1 0. In the present paper we extend Algorthms 1.2 and 1.3 to general reflexve Banach spaces usng a well chosen convex functon f. More precsely, we ntroduce the followng algorthm: x 0 X, 1.4 y n = Res f λ na x n, C n = {z X : D f z, y n D f z, x n }, Q n = {z X : f x 0 f x n, z x n 0}, x n+1 = proj f C n Q n x 0, n = 0, 1, 2,..., where {λ n } n N s a gven sequence of postve real numbers, Res f A s the resolvent of A relatve to f, ntroduced and studed n [4], f s the gradent of f and proj f C s the Bregman projecton of X onto C nduced by f see Secton 2.4. Algorthm 1.4 s more flexble than 1.3 because t leaves us the freedom of fttng the functon f to the nature of the operator A especally when A s the subdfferental of some functon and of the space X n ways whch make the applcaton of 1.4 smpler than that of 1.3. It should be observed that f X s a Hlbert space H, then usng n 1.4 the functon f x = 1/2 x 2, one obtans exactly Algorthm 1.2. If X s not a Hlbert space, but stll a unformly convex and unformly smooth Banach space X, then settng f x = 1/2 x 2 n 1.4, one obtans exactly 1.3. We also note that the choce f x = 1/2 x 2 n some Banach spaces may make the computatons n Algorthm 1.3 qute dffcult. These computatons can be smplfed by an approprate choce of f. For nstance, f X = l p or X = L p wth p 1,, and f x = 1/p x p n 1.4, then the computatons become smpler than those requred n 1.3, whch corresponds to f x = 1/2 x 2. As a matter of fact, we propose two extensons of Algorthm 1.4 see Algorthms 4.1 and 4.4 whch approxmate a common zero of several maxmal monotone operators and whch allow computatonal errors. These algorthms are smlar to but dfferent

4 4 SIMEON REICH AND SHOHAM SABACH from the one we have recently studed n [34]. They also dffer from the algorthm n [6] n the defnton of the sets C n and n our takng nto account possble computatonal errors. Our man results Theorems 1 and 2 are formulated and proved n Secton 4. The next secton s devoted to several prelmnary defntons and results. In secton 3 we prove two auxlary results whch are used n the proofs of our man results n Secton 4. The behavor of Algorthm 1.4 when the operator A s zero free s analyzed n Secton 5 see Theorem 3. The sxth secton contans three corollares of Theorems 1, 2 and 3. In the seventh and last secton we present an applcaton of Theorems 1, 2 and Prelmnares 2.1. Some facts about Legendre functons. Legendre functons mappng a general Banach space X nto, + ] are defned n [3]. Accordng to [3, Theorems 5.4 and 5.6], snce X reflexve, the functon f s Legendre f and only f t satsfes the followng two condtons: L1 The nteror of the doman of f, nt dom f, s nonempty, f s Gâteaux dfferentable see below on nt dom f, and dom f = nt dom f; L2 The nteror of the doman of f, nt dom f, s nonempty, f s Gâteaux dfferentable on nt dom f, and dom f = nt dom f. Snce X s reflexve, we always have f 1 = f see [8, p. 83]. Ths fact, when combned wth condtons L1 and L2, mples the followng equaltes: f = f 1, ran f = dom f = nt dom f and ran f = dom f = nt dom f. Also, condtons L1 and L2, n conjuncton wth [3, Theorem 5.4], mply that the functons f and f are strctly convex on the nteror of ther respectve domans. Several nterestng examples of Legendre functons are presented n [2] and [3]. Among them are the functons 1 s s wth s 1,, where the Banach space X s smooth and strctly convex and, n partcular, a Hlbert space. The functon f s called cofnte f dom f = X A property of gradents. For any convex f : X, + ] we denote by dom f the set {x X : f x < + }. For any x dom f and y X, we

5 STRONG CONVERGENCE THEOREMS 5 denote by f x, y the rght-hand dervatve of f at x n the drecton y, that s, f fx + ty fx x, y := lm. t 0 t The functon f s sad to be Gâteaux dfferentable at x f lm t 0 fx + ty fx /t exsts for any y. The functon f s sad to be Fréchet dfferentable at x f ths lmt s attaned unformly n y = 1. Fnally, f s sad to be unformly Fréchet dfferentable on a subset E of X f the lmt s attaned unformly for x E and y = 1. We wll need the followng result. Proposton 1 cf. [34, Proposton 2]. If f : X R s unformly Fréchet dfferentable and bounded on bounded subsets of X, then f s unformly contnuous on bounded subsets of X from the strong topology of X to the strong topology of X Some facts about totally convex functons. Let f : X, + ] be convex. The functon D f : dom f nt dom f [0, + ], defned by 2.1 D f y, x := fy fx f x, y x, s called the Bregman dstance wth respect to f cf. [18]. If f s a Gâteaux dfferentable functon, then the Bregman dstance has the followng mportant property, called the three pont dentty: for any x, y, z nt dom f, 2.2 D f x, y + D f y, z D f x, z = fz fy, x y. Recall that, accordng to [13, Secton 1.2, p. 17] see also [12], the functon f s called totally convex at a pont x nt dom f f ts modulus of total convexty at x, that s, the functon υ f : nt dom f [0, + [0, + ], defned by 2.3 υ f x, t := nf {D f y, x : y dom f, y x = t}, s postve whenever t > 0. The functon f s called totally convex when t s totally convex at every pont x nt dom f. In addton, the functon f s called totally convex on bounded sets f υ f E, t s postve for any nonempty bounded subset E of X and for any t > 0, where the modulus of total convexty of the functon f on the set E s the functon υ f : nt dom f [0, + [0, + ] defned by υ f E, t := nf {υ f x, t x E dom f}. Examples of totally convex functons can be found, for example, n [13, 17]. The followng proposton summarzes some propertes of the modulus of total convexty. Proposton 2 cf. [13, Proposton 1.2.2, p. 18]. Let f be a proper, convex and lower semcontnuous functon. If x nt dom f, then The doman of υ f x, s an nterval of the form [0, τ f x or [0, τ f x] wth τ f x 0, + ].

6 6 SIMEON REICH AND SHOHAM SABACH If c [1, + and t 0, then υ f x, ct cυ f x, t. The functon υ f x, s superaddtve, that s, for any s, t [0, +, we have υ f x, s + t υ f x, s + υ f x, t. v The functon υ f x, s ncreasng; t s strctly ncreasng f and only f f s totally convex at x. The followng proposton follows from [15, Proposton 2.3, p. 39] and [44, Theorem , p. 164]. cofnte. results. Proposton 3. If f s Fréchet dfferentable and totally convex, then f s The next proposton turns out to be very useful n the proof of our man Proposton 4 cf. [36, Proposton 2.2, p. 3]. If x dom f, then the followng statements are equvalent: The functon f s totally convex at x; For any sequence {y n } n N dom f, lm D f y n, x = 0 lm y n x = 0. n + n + Recall that the functon f s called sequentally consstent see [17] f for any two sequences {x n } n N and {y n } n N n X such that the frst one s bounded, lm D f y n, x n = 0 lm y n x n = 0. n + n + Proposton 5 cf. [13, Lemma 2.1.2, p. 67]. If dom f contans at least two ponts, then the functon f s totally convex on bounded sets f and only f the functon f s sequentally consstent The resolvent of A relatve to f. Let A : X 2 X be an operator and assume that f Gâteaux dfferentable. The operator Prt f A := f + A 1 : X 2 X s called the protoresolvent of A, or, more precsely, the protoresolvent of A relatve to f. Ths allows us to defne the resolvent of A, or, more precsely, the resolvent of A relatve to f, ntroduced and studed n [4], as the operator Res f A : X 2X gven by Res f A := Prtf A f. Ths operator s sngle-valued when A s monotone and f s strctly convex on nt dom f. If A = ϕ, where ϕ s a proper, lower semcontnuous and convex functon, then we denote Prox f ϕ := Prt f ϕ and proxf ϕ := Res f ϕ.

7 STRONG CONVERGENCE THEOREMS 7 If C s a nonempty, closed and convex subset of X, then the ndcator functon ι C of C, that s, the functon ι C x := { 0 f x C + f x / C s proper, convex and lower semcontnuous, and therefore ι C exsts and s a maxmal monotone operator wth doman C. The operator prox f ι C s called the Bregman projecton onto C wth respect to f cf. [9] and we denote t by proj f C. Note that f X s a Hlbert space and fx = 1 2 x 2, then the Bregman projecton of x onto C,.e., argmn { y x : y C}, s the metrc projecton P C. Recall that the Bregman projecton of x onto the nonempty, closed and convex set K dom f s the necessarly unque vector proj f K x K satsfyng D f proj f K x, x = nf {D f y, x : y K}. Smlarly to the metrc projecton n Hlbert spaces, Bregman projectons wth respect to totally convex and dfferentable functons have a varatonal characterzaton. Proposton 6 cf. [17, Corollary 4.4, p. 23]. Suppose that f s totally convex on nt dom f. Let x nt dom f and let K nt dom f be a nonempty, closed and convex set. If ˆx K, then the followng condtons are equvalent: The vector ˆx s the Bregman projecton of x onto K wth respect to f; The vector ˆx s the unque soluton of the varatonal nequalty f x f z, z y 0, y K; The vector ˆx s the unque soluton of the nequalty D f y, z + D f z, x D f y, x, y K. For the next techncal result we need to defne, for any λ > 0, the Yosda approxmaton of A by A λ = f f Res f λa /λ. We have the followng propertes of the Yosda approxmaton A λ. Proposton 7: For any λ > 0 and for any x X, we have Res f λa x, A λ x graph A; 0 Ax f and only f 0 A λ x.

8 8 SIMEON REICH AND SHOHAM SABACH Proof. Indeed, Res f λa x = f + λa 1 f x f x f + λa Res f λa x f f Res f λa x /λ A Res fλa x A λ x A Res fλa x. Indeed, 0 Ax 0 λax f x f + λa x x f + λa 1 f x f x f Res fλa x 0 f f Res f λa x 0 λa λ x 0 A λ x. Now we can prove the followng mportant property of the resolvent. Proposton 8: Let A : X 2 X be a maxmal monotone operator such that A 1 0. Then D f u, Res f λa x + D f Res f λa x, x D f u, x for all λ > 0, u A 1 0 and x X. Proof. Let λ > 0, u A 1 0 and x X be gven. By the monotoncty of A, the three pont dentty 2.2 and Proposton 7, we have D f u, x = D f u, Res f λa x + D f Res f λa x, x + f Res fλa x f x, u ResfλA x = D f u, Res f λa x + D f Res f λa x, x + λ A λ x, u Res fλa x D f u, Res f λa x + D f Res f λa x, x. 3. Auxlary Results In ths secton we prove two lemmata whch are used n the proofs of our man results n Secton 4. Lemma 1: Let f : X R be a totally convex functon. If {D f x n, x 0 } n N s bounded, then the sequence {x n } n N s bounded too. Proof. Snce the sequence {D f x n, x 0 } n N s bounded, there exsts M > 0 such that D f x n, x 0 < M for any n N. Therefore the sequence {ν f x 0, x n x 0 } n N s bounded by M too, because from the defnton of the

9 STRONG CONVERGENCE THEOREMS 9 modulus of total convexty see 2.3 we get that 3.1 ν f x 0, x n x 0 D f x n, x 0 M. Snce the functon f s totally convex, the functon ν f x, s strctly ncreasng and postve on 0, cf. Proposton 2v. Ths mples, n partcular, that ν f x, 1 > 0 for all x X. Now suppose by way of contradcton that the sequence {x n } n N s not bounded. Then there exsts a sequence {n k } k N of postve real numbers such that lm x n k = +. k + Consequently, lm k + x nk x 0 = +. Ths shows that the sequence {ν f x 0, x n x 0 } n N s not bounded. Indeed, there exsts some k 0 > 0 such that x nk x 0 > 1 for any k > k 0 and then, by Proposton 2, we see that ν f x 0, x nk x 0 x nk x 0 ν f x 0, 1 +, because, as noted above, ν f x 0, 1 > 0. Ths contradcts 3.1. Hence the sequence {x n } n N s ndeed bounded, as clamed. Lemma 2: Let f : X R be a totally convex functon and let C be a nonempty, closed and convex subset of X. Suppose that the sequence {x n } n N s bounded and any weak subsequental lmt of {x n } n N belongs to C. If D f x n, x 0 D f proj f C x 0, x 0 for any n N, then {x n } n N converges strongly to proj f C x 0. Proof. Denote proj f C x 0 = ũ. The three pont dentty see 2.2 and the assumpton D f x n, x 0 D f ũ, x 0 yelds D f x n, ũ = D f x n, x 0 + D f x 0, ũ fũ fx 0, x n x 0 D f ũ, x 0 + D f x 0, ũ fũ fx 0, x n x 0 = fũ fx 0, ũ x 0 fũ fx 0, x n x 0 = fũ fx 0, ũ x n. Snce {x n } n N s bounded there s a weakly convergent subsequence {x n } N and denote ts weak lmt by v. We know that v C. It follows from Proposton 6 that lm sup + D f x n, ũ lm sup fũ fx 0, ũ x n + = fũ fx 0, ũ v 0. Hence lm D f x n, ũ = 0. +

10 10 SIMEON REICH AND SHOHAM SABACH Proposton 4 now mples that x n ũ. It follows that the whole sequence {x n } n N converges strongly to ũ = proj f C x 0, as clamed Two Strong Convergence Theorems In ths secton we study the followng algorthm when Z := N x 0 X, ηn = ξn + 1 λ fy n n fx n, ξn A yn, w n = f λ nη n + fx n, C n = { z X : D f z, y n Df z, w n }, C n := N =1 C n, Q n = {z X : fx 0 fx n, z x n 0}, x n+1 = proj f C n Q n x 0, n = 0, 1, 2,..., =1 A 1 0 : Theorem 1: Let A : X 2 X, = 1, 2,..., N, be N maxmal monotone operators such that Z := N =1 A 1 0. Let f : X R be a Legendre functon whch s bounded, unformly Fréchet dfferentable and totally convex on bounded subsets of X. Assume further that f s bounded and unformly Fréchet dfferentable on bounded subsets of X. Then, for each x 0 X, there are sequences {x n } n N whch satsfy 4.1. If, for each = 1, 2,..., N, lm nf n + λ n > 0, and the sequences of errors { } ηn n N X satsfy lm n + ηn = 0, then each such sequence {x n } n N converges strongly to proj f Z x 0 as n +. Proof. Note that dom f = X because dom f = X and f s Legendre. Hence t follows from [4, Corollary 3.14, p. 606] that dom Res f λa = X. We begn wth the followng clam. Clam 1: There are sequences {x n } n N whch satsfy 4.1. As a matter of fact, we wll prove that, for each x 0 X, there exsts a sequence {x n } n N whch s generated by 4.1 wth η n = 0 for all = 1, 2,..., N and n N. It s obvous that C n are closed and convex sets for any = 1, 2,..., N. Hence C n s also closed and convex. It s also obvous that Q n s a closed and convex set. Let u Z. For any n N we have from Proposton 8 that D f u, y n = Df u, Res f λ n Aw n D f u, w n, whch mples that u C n. Snce ths holds for any = 1, 2,..., N, t follows that u C n. Thus Z C n for any n N. On the other hand t s obvous that Z Q 0 = X. Thus Z C 0 Q 0, and therefore x 1 = proj f C 0 Q 0 x 0 s well defned. Now suppose that Z C n 1 Q n 1 for some n 1. Then t follows that there exsts x n C n 1 Q n 1 such that x n = proj f C n 1 Q n 1 x 0 snce C n 1 Q n 1 s

11 STRONG CONVERGENCE THEOREMS 11 a nonempty, closed and convex subset of X. So from Proposton 6 we have fx 0 fx n, y x n 0, for any y C n Q n. Hence we obtan that Z Q n. Therefore Z C n Q n and hence x n+1 = proj f C n Q n x 0 s well defned. Consequently, we see that Z C n Q n for any n N. Thus the sequence we constructed s ndeed well defned and satsfes 4.1, as clamed. From now on we fx an arbtrary sequence {x n } n N satsfyng 4.1. It s clear from the proof of Clam 1 that Z C n Q n for each n N. Clam 2: The sequence {x n } n N s bounded. It follows from the defnton of Q n and Proposton 6 that proj f Q n x 0 = x n. Furthermore, by Proposton 6, for each u Z, we have 4.2 D f x n, x 0 = D f proj f Q n x 0, x 0 D f u, x 0 D f u, proj f Q n x 0 D f u, x 0. Hence the sequence {D f x n, x 0 } n N s bounded by D f u, x 0 for any u Z. Therefore by Lemma 1 the sequence {x n } n N s bounded too, as clamed. Clam 3: Every weak subsequental lmt of {x n } n N belongs to Z. It follows from the defnton of Q n and Proposton 6 that proj f Q n x 0 = x n. Snce x n+1 Q n, t follows from Proposton 6 that D f x n+1, proj f Q n x 0 + D f proj fqn x 0, x 0 D f x n+1, x 0 and hence 4.3 D f x n+1, x n + D f x n, x 0 D f x n+1, x 0. Therefore the sequence {D f x n, x 0 } n N s ncreasng and snce t s also bounded see Clam 2, lm n + D f x n, x 0 exsts. Thus from 4.3 t follows that lm D f x n+1, x n = 0. n + Proposton 5 now mples that lm n + x n+1 x n = 0. Snce wn = f λ nηn + fx n and f s unformly contnuous on bounded subsets of X by Proposton 1, t follows that lm w n x n = 0 n +

12 12 SIMEON REICH AND SHOHAM SABACH for any = 1, 2,..., N, and hence lm n + D f xn, wn = 0. For any = 1, 2,..., N, the three pont dentty see 2.2 mples that D f xn+1, w n = Df x n+1, x n D f xn, w n + fxn fw n, x n+1 x n. Therefore lm D f xn+1, w n = 0. n + Next, for any = 1, 2,..., N, t follows from the ncluson x n+1 C n that D f xn+1, y n Df xn+1, w n. Hence lm n + D f xn+1, y n = 0. Proposton 5 now mples that lm n + y n x n+1 = 0. Therefore, for any = 1, 2,..., N, we have y n x n y n x n+1 + xn+1 x n 0. Ths means that the sequence { } yn s bounded for any = 1, 2,..., N. Now n N let { x nj }j N be a weakly convergent subsequence of {x n} n N and denote ts weak } lmt by v. Then {y nj also converges weakly to v for any = 1, 2,..., N. Snce j N lm nf n + λ n > 0 and lm n + η n = 0, t follows from Proposton 1 that ξn = 1 fxn λ fyn + ηn 0 n for any = 1, 2,..., N. Snce ξn Ayn and A s monotone, t follows that η ξ n, z yn 0 for all z, η graph A. Ths, n turn, mples that η, z v 0 for all z, η graph A. Therefore, usng the maxmal monotoncty of A, we now obtan that v A 1 0 for each = 1, 2,..., N. Thus v Z and ths proves Clam 3. Clam 4: The sequence {x n } n N converges strongly to proj f Z x 0. Let ũ = proj f Z x 0. Snce x n+1 = proj f C n Q n x 0 and Z s contaned n C n Q n, we have D f x n+1, x 0 D f ũ, x 0. Therefore Lemma 2 mples that {x n } n N converges strongly to ũ = proj f Z x 0, as clamed. Ths completes the proof of Theorem 1. We now present another result whch s smlar to Theorem 1, but wth a dfferent type of errors. More precsely, we study the followng algorthm when

13 STRONG CONVERGENCE THEOREMS 13 Z := N =1 A : x 0 X, yn = Res f λ Ax n + e n, n Cn = { z X : D f z, y n Df z, xn + en}, C n := N =1 C n, Q n = {z X : fx 0 fx n, z x n 0}, x n+1 = proj f H n W n x 0, n = 0, 1, 2,...,. Theorem 2: Let A : X 2 X, = 1, 2,..., N, be N maxmal monotone operators such that Z := N =1 A 1 0. Let f : X R be a Legendre functon whch s bounded, unformly Fréchet dfferentable and totally convex on bounded subsets of X. Then, for each x 0 X, there are sequences {x n } n N whch satsfy 4.4. If, for each = 1, 2,..., N, lm nf n + λ n > 0, and the sequences of errors { } e n X satsfy lm n N n + e n = 0, then each such sequence {x n } n N converges strongly to proj f Z x 0 as n +. Proof. Note that dom f = X because dom f = X and f s Legendre. Hence t follows from [4, Corollary 3.14, p. 606] that dom Res f λa = X. We begn wth the followng clam. Clam 1: There are sequences {x n } n N whch satsfy 4.4. As a matter of fact, we wll prove that, for each x 0 X, there exsts a sequence {x n } n N whch s generated by 4.4 wth e n = 0 for all = 1, 2,..., N and n N. It s obvous that C n are closed and convex sets for any = 1, 2,..., N. Hence C n s also closed and convex. It s also obvous that Q n s a closed and convex set. Let u Z. For any n N, we obtan from Proposton 8 that D f u, y n = Df u, Res f λ Ax n + e n D f u, xn + e n, n whch mples that u C n. Snce ths holds for any = 1, 2,..., N, t follows that u C n. Thus Z C n for any n N. On the other hand t s obvous that Z Q 0 = X. Thus Z C 0 Q 0, and therefore x 1 = proj f C 0 Q 0 x 0 s well defned. Now suppose that Z C n 1 Q n 1 for some n 1. The t follows that there exsts x n C n 1 Q n 1 such that x n = proj f C n 1 Q n 1 x 0 snce C n 1 Q n 1 s a nonempty, closed and convex subset of X. So from Proposton 6 we have fx 0 fx n, y x n 0, for any y C n Q n. Hence we obtan that Z Q n. Therefore Z C n Q n and hence x n+1 = proj f C n Q n x 0 s well defned. Consequently, we see that Z C n Q n for any n N. Thus the sequence we constructed s ndeed well defned and satsfes 4.4, as clamed.

14 14 SIMEON REICH AND SHOHAM SABACH From now on we fx an arbtrary sequence {x n } n N satsfyng 4.4. It s clear from the proof of Clam 1 that Z C n Q n for each n N. Clam 2: The sequence {x n } n N s bounded. It follows from the defnton of Q n and Proposton 6 that proj f Q n x 0 = x n. Furthermore, by Proposton 6, for each u Z, we have 4.5 D f x n, x 0 = D f proj f Q n x 0, x 0 D f u, x 0 D f u, proj f Q n x 0 D f u, x 0. Hence the sequence {D f x n, x 0 } n N s bounded by D f u, x 0 for any u Z. Therefore by Lemma 1 the sequence {x n } n N s bounded too, as clamed. Clam 3: Every weak subsequental lmt of {x n } n N belongs to Z. It follows from the defnton of Q n and Proposton 6 that proj f Q n x 0 = x n. Snce x n+1 Q n, t follows from Proposton 6 that D f x n+1, proj f Q n x 0 + D f proj fqn x 0, x 0 D f x n+1, x 0 and hence 4.6 D f x n+1, x n + D f x n, x 0 D f x n+1, x 0. Therefore the sequence {D f x n, x 0 } n N s ncreasng and snce t s also bounded see Clam 2, lm n + D f x n, x 0 exsts. Thus from 4.6 t follows that 4.7 lm n + D f x n+1, x n = 0. Proposton 5 now mples that lm n + x n+1 x n = 0. For any = 1, 2,..., N, t follows from the defnton of the Bregman dstance see 2.1 that D f xn, x n + e n = f xn f x n + e n fxn + e n, x n x n + e n = f x n f x n + e n + fxn + e n, e n. The functon f s bounded on bounded subsets of X and therefore f s bounded on bounded subsets of X see [13, Proposton , p. 17]. In addton, f s unformly Fréchet dfferentable and therefore f s unformly contnuous on bounded subsets see [1, Theorem 1.8, p. 13]. Hence, snce lm n + e n = 0, t follows that 4.8 lm D f xn, x n + e n = 0. n +

15 STRONG CONVERGENCE THEOREMS 15 For any = 1, 2,..., N, t follows from the three pont dentty see 2.2 that D f xn+1, x n + e n = Df x n+1, x n + D f xn, x n + e n + fx n fx n + e n, x n+1 x n. Snce lm n + x n+1 x n = 0 and f s bounded on bounded subsets of X, 4.7 and 4.8 mply that lm D f xn+1, x n + e n = 0. n + For any = 1, 2,..., N, t follows from the ncluson x n+1 C n that D f xn+1, y n Df xn+1, x n + e n. Hence lm n + D f xn+1, y n = 0. Proposton 5 now mples that lm n + y n x n+1 = 0. Therefore, for any = 1, 2,..., N, we have y n x n y n x n+1 + xn+1 x n 0. Ths means that the sequence { } yn s bounded for any = 1, 2,..., N. Now n N let { x nj }j N be a weakly convergent subsequence of {x n} n N and denote ts weak } lmt by v. Then {y nj also converges weakly to v for any = 1, 2,..., N. j N Let ξ n Ay n, snce lm nf n + λ n > 0 and lm n + e n = 0, t follows from Proposton 1 that ξn = 1 fxn λ + e n fyn 0 n for any = 1, 2,..., N. Snce ξn Ayn and A s monotone, t also follows that η ξ n, z yn 0 for all z, η graph A. Ths, n turn, mples that η, z v 0 for all z, η graph A. Therefore, usng the maxmal monotoncty of A, we now obtan that v A 1 0 for each = 1, 2,..., N. Thus v Z and ths proves Clam 3. Clam 4: The sequence {x n } n N converges strongly to proj f Z x 0. Let ũ = proj f Z x 0. Snce x n+1 = proj f C n Q n x 0 and Z s contaned n C n Q n, we have D f x n+1, x 0 D f ũ, x 0. Therefore Lemma 2 mples that {x n } n N converges strongly to ũ = proj f Z x 0, as clamed. Ths completes the proof of Theorem 2.

16 16 SIMEON REICH AND SHOHAM SABACH 5. Zero Free Operators Ths secton concerns the case where our two algorthms are appled to a sngle zero free operator A. In ths case both our algorthms take the form 5.1 and 5.2 x 0 X, η n = ξ n + 1 λ n fy n fx n, ξ n Ay n, w n = f λ n η n + fx n, C n = {z X : D f z, y n D f z, x n }, Q n = {z X : fx 0 fx n, z x n 0}, x n+1 = proj f C n Q n x 0, n = 0, 1, 2,..., x 0 X, y n = Res f λ na x n + e n, C n = {z X : D f z, y n D f z, x n + e n }, Q n = {z X : fx 0 fx n, z x n 0}, x n+1 = proj f C n Q n x 0, n = 0, 1, 2,...,. We frst recall the followng lemma see [34, Lemma 1]: Lemma 3: If A : X 2 X s a maxmal monotone operator wth bounded doman, then A 1 0. Now we can prove that the generaton of an nfnte sequence by Algorthm 5.1 or 5.2 does not depend on the zero set A 1 0 of A beng not empty. Theorem 3. Let A : X 2 X be a maxmal monotone operator. Let f : X R be a Legendre functon whch s bounded, unformly Fréchet dfferentable and totally convex on bounded subsets of X. In case of Algorthm 5.1 assume, n addton, that f s bounded and unformly Fréchet dfferentable on bounded subsets of X. Then, for each x 0 X, there are sequences {x n } n N whch satsfy ether 5.1 or 5.2. If lm nf n + λ n > 0, and ether the sequence of errors {η n } n N X satsfes lm n + η n = 0 or the sequence of errors {e n } n N X satsfes lm n + e n = 0, then ether A 1 0 and each such sequence {x n } n N converges strongly to proj f A 1 0 x 0 or A 1 0 = and each such sequence {x n } n N satsfes lm n + x n = +. Proof. In vew of Theorem 1 and Theorem 2, we only need to consder the case where A 1 0 =. Frst of all we prove that n ths case, for each x 0 X, there s a sequence {x n } n N whch satsfes ether 5.1 wth η n = 0 or 5.2 wth e n = 0 for all n N.

17 STRONG CONVERGENCE THEOREMS 17 We prove ths by nducton. We frst check that the ntal step n = 0 s well defned. Indeed, the problem 0 Ax + 1 λ 0 fx fx 0 always has a soluton y 0, ξ 0 because t s equvalent to the problem x =Res f λ 0A x 0 and ths problem does have a soluton snce dom Res f λa = X see Proposton 3 and [4, Theorem 3.13v, p. 606]. Now note that Q 0 = X. Snce C 0 cannot be empty y 0 C 0, the next terate x 1 can be generated; t s the Bregman projecton of x 0 onto C 0 = Q 0 C 0. Note that whenever x n s generated, y n and ξ n can further be obtaned because the proxmal subproblems always have solutons. Suppose now that x n and y n, ξ n have already been defned for n = 0,..., ˆn. We have to prove that xˆn+1 s also well defned. To ths end, take any z 0 dom A and defne and ρ = max { y n z 0 : n = 0,..., ˆn} hx = { 0, x z 0 ρ + 1 +, otherwse. Then h : X, + ] s a proper, convex and lower semcontnuous functon, ts subdfferental h s maxmal monotone see [31, Theorem 2.13, p. 124], and A = A + h s also maxmal monotone see [37]. Furthermore, A z = A z for all z z 0 < ρ + 1. Therefore ξ n A y n for n = 0,..., ˆn. We conclude that x n and y n, ξ n also satsfy the condtons of Theorems 1 and 2 appled to the problem 0 A x. Snce A has a bounded effectve doman, ths problem has a soluton by Lemma 3. Thus t follows from Clam 1 n the proofs of Theorems 1 and 2 that xˆn+1 s well defned n both Algorthms 5.1 and 5.2. Hence the whole sequence {x n } n N s well defned, as asserted. If {x n } n N were to have a bounded subsequence, then t would follow from Clam 3 n the proofs of Theorems 1 and 2 that A had a zero. Therefore f A 1 0 =, then lm n + x n = +, as asserted. 6. Consequences of the Strong Convergence Theorems Algorthm 1.4 s a specal case of Algorthm 5.1 when η n = 0 for all n N, and a specal case of Algorthm 5.2 when e n = 0 for all n N. Hence as a drect consequence of Theorems 1, 2 and 3 we obtan the followng result:

18 18 SIMEON REICH AND SHOHAM SABACH Corollary 1. Let A : X 2 X be a maxmal monotone operator. Let f : X R be a Legendre functon whch s bounded, unformly Fréchet dfferentable and totally convex on bounded subsets of X, and suppose that lm nf n + λ n > 0. Then for each x 0 X, the sequence {x n } n N generated by 1.4 s well defned, and ether A 1 0 and {x n } n N converges strongly to proj f A 1 0 x 0 as n +, or A 1 0 = and lm n + x n = +. Notable corollares of Theorems 1, 2 and 3 occur when the space X s both unformly smooth and unformly convex. In ths case the functon fx = 1 2 x 2 s Legendre cf. [3, Lemma 6.2, p. 24] and unformly Fréchet dfferentable on bounded subsets of X. Accordng to [14, Corollary 1, p. 325], f s sequentally consstent snce X s unformly convex and hence f s totally convex on bounded subsets of X. Therefore Theorems 1, 2 and 3 hold n ths context and lead us to the followng two results whch, n some sense, complement Theorem 3.1 n [42] see also Theorem 3.5 n [29]. Corollary 2. Let X be a unformly smooth and unformly convex Banach space and let A : X 2 X be a maxmal monotone operator. Then, for each x 0 X, the sequence {x n } n N generated by 1.3 s well defned. If lm nf n + λ n > 0, then ether A 1 0 and {x n } n N converges strongly to Q A 1 0 x 0 as n +, or A 1 0 = and lm n + x n = +. Corollary 3. Let X be a Hlbert space and let A : X 2 X be a maxmal monotone operator. Then, for each x 0 X, the sequence {x n } n N generated by 1.2 s well defned. If lm nf n + λ n > 0, then ether A 1 0 and {x n } n N converges strongly to P A 1 0x 0 as n +, or A 1 0 = and lm n + x n = +. These corollares also hold n the presence of computatonal errors as n Theorems 1, 2 and An Applcaton of the Strong Convergence Theorems Let g : X, + ] be a proper, convex and lower semcontnuous functon. Recall that the subdfferental g of g s defned for any x X by g x := {ξ X : ξ, y x g y g x y X}. Applyng Theorems 1, 2 and 3 to the subdfferental of g, we obtan an algorthm for fndng mnmzers of g. Proposton 9. Let g : X, + ] be a proper, convex and lower semcontnuous functon whch attans ts mnmum over X. If f : X R s a Legendre functon whch s bounded, unformly Fréchet dfferentable, and totally convex on bounded subsets of X, and {λ n } n N s a postve sequence wth lm nf n + λ n > 0,

19 STRONG CONVERGENCE THEOREMS 19 then, for each x 0 X, the sequence {x n } n N generated by x 0 X, 0 = ξ n + 1 λ n fy n fx n, ξ n g y n, C n = {z X : D f z, y n D f z, x n }, Q n = {z X : fx 0 fx n, z x n 0}, x n+1 = proj f C n Q n x 0, n = 0, 1, 2,..., converges strongly to a mnmzer of g as n +. If g does not attan ts mnmum over X, then lm n + x n = +. Proof. The subdfferental g of g s a maxmal monotone operator because g s a proper, convex and lower semcontnuous functon see [31, Theorem 2.13, p. 124]. Snce the zero set of g concdes wth the set of mnmzers of g, Proposton 9 follows mmedately from Theorems 1, 2 and 3. Note that n ths case s equvalent to y n = arg mn x X { g x + 1 } D f x, x n λ n 0 g y n + 1 λ n fy n fx n. 8. Acknowledgements The frst author was partally supported by the Israel Scence Foundaton Grant 647/07, by the Fund for the Promoton of Research at the Technon and by the Technon Presdent s Research Fund. Both authors thank the referee for several helpful comments. References [1] Ambrosett, A. and Prod, G.: A prmer of nonlnear analyss, Cambrdge Unversty Press, Cambrdge, [2] Bauschke, H. H. and Borwen, J. M.: Legendre functons and the method of random Bregman projectons, J. Convex Anal , [3] Bauschke, H. H., Borwen, J. M. and Combettes, P. L.: Essental smoothness, essental strct convexty, and Legendre functons n Banach spaces, Commun. Contemp. Math , [4] Bauschke, H. H., Borwen, J. M. and Combettes, P. L.: Bregman monotone optmzaton algorthms, SIAM J. Control Optm , [5] Bauschke, H. H. and Combettes, P. L.: A weak-to-strong convergence prncple for Fejérmonotone methods n Hlbert spaces, Math. Oper. Res , [6] Bauschke, H. H. and Combettes, P. L.: Constructon of best Bregman approxmatons n reflexve Banach spaces, Proc. Amer. Math. Soc ,

20 20 SIMEON REICH AND SHOHAM SABACH [7] Bauschke, H. H., Matoušková, E. and Rech, S.: Projecton and proxmal pont methods: convergence results and counterexamples, Nonlnear Anal , [8] Bonnans, J. F. and Shapro, A.: Perturbaton analyss of optmzaton problems, Sprnger Verlag, New York, [9] Bregman, L. M.: A relaxaton method for fndng the common pont of convex sets and ts applcaton to the soluton of problems n convex programmng, USSR Comput. Math. and Math. Phys , [10] Brézs, H. and Lons, P.-L.: Produts nfns de résolvantes, Israel J. Math , [11] Bruck, R. E. and Rech, S.: Nonexpansve projectons and resolvents of accretve operators, Houston J. Math , [12] Butnaru, D., Censor, Y. and Rech, S.: Iteratve averagng of entropc projectons for solvng stochastc convex feasblty problems, Computatonal Optmzaton and Applcatons , [13] Butnaru, D. and Iusem, A. N.: Totally convex functons for fxed ponts computaton and nfnte dmensonal optmzaton, Kluwer Academc Publshers, Dordrecht, [14] Butnaru, D., Iusem, A. N. and Resmerta, E.: Total convexty for powers of the norm n unformly convex Banach spaces, J. Convex Anal , [15] Butnaru, D., Iusem, A. N. and Zălnescu, C.: On unform convexty, total convexty and convergence of the proxmal pont and outer Bregman projecton algorthms n Banach spaces, J. Convex Anal , [16] Butnaru, D. and Kassay, G.: A proxmal-projecton method for fndng zeroes of set-valued operators, SIAM J. Control Optm , [17] Butnaru, D. and Resmerta, E.: Bregman dstances, totally convex functons and a method for solvng operator equatons n Banach spaces, Abstr. Appl. Anal. 2006, Art. ID 84919, [18] Censor, Y. and Lent, A.: An teratve row-acton method for nterval convex programmng, J. Optm. Theory Appl , [19] Censor, Y. and Zenos, S. A.: Proxmal mnmzaton algorthm wth D-functons, J. Optm. Theory Appl , [20] Coranescu, I.: Geometry of Banach spaces, dualty mappngs and nonlnear problems, Kluwer Academc Publshers, Dordrecht, [21] Combettes, P. L.: Strong convergence of block-teratve outer approxmaton methods for convex optmzaton, SIAM J. Control Optm , [22] Ecksten, J.: Nonlnear proxmal pont algorthms usng Bregman functons, wth applcaton to convex programmng, Math. Oper. Res , [23] Gárcga Otero, R. and Svater, B. F.: A strongly convergent hybrd proxmal method n Banach spaces, J. Math. Anal. Appl , [24] Güler, O.: On the convergence of the proxmal pont algorthm for convex mnmzaton, SIAM J. Control Optm , [25] Kammura, S. and Takahash, W.: Approxmatng solutons of maxmal monotone operators n Hlbert spaces, J. Approx. Theory , [26] Kammura, S. and Takahash, W.: Weak and strong convergence of solutons to accretve operator nclusons and applcatons, Set-Valued Anal , [27] Kammura, S. and Takahash, W.: Strong convergence of a proxmal-type algorthm n a Banach space, SIAM J. Optm ,

21 STRONG CONVERGENCE THEOREMS 21 [28] Martnet, B.: Régularsaton d néquatons varatonelles par approxmatons successves, Revue Françase d Informatque et de Recherche Opératonelle , [29] Nakajo, K. and Takahash, W.: Strong convergence theorems for nonexpansve mappngs and nonexpansve semgroups, J. Math. Anal. Appl , [30] Nevanlnna, O. and Rech, S.: Strong convergence of contracton semgroups and of teratve methods for accretve operators n Banach spaces, Israel J. Math , [31] Pascal, D. and Sburlan, S.: Nonlnear mappngs of monotone type, Sjthoff & Nordhoff Internatonal Publshers, Alphen aan den Rjn, [32] Rech, S.: Weak convergence theorems for nonexpansve mappngs n Banach spaces, J. Math. Anal. Appl , [33] Rech, S.: A weak convergence theorem for the alternatng method wth Bregman dstances, n Theory and applcatons of nonlnear operators of accretve and monotone type, Marcel Dekker, New York, 1996, [34] Rech, S. and Sabach, S.: A strong convergence theorem for a proxmal-type algorthm n reflexve Banach spaces, preprnt, [35] Rech, S. and Zaslavsk, A. J.: Infnte products of resolvents of accretve operator, Topol. Methods Nonlnear Anal , [36] Resmerta, E.: On total convexty, Bregman projectons and stablty n Banach spaces, J. Convex Anal , [37] Rockafellar, R. T.: On the maxmalty of sums of nonlnear monotone operators, Trans. Amer. Math. Soc , [38] Rockafellar, R. T.: Monotone operators and the proxmal pont algorthm, SIAM J. Control Optm , [39] Rockafellar, R. T.: Augmented Lagrangans and applcatons of the proxmal pont algorthm n convex programmng, Math. Oper. Res , [40] Rockafellar, R. T. and Wets, R. J.-B.: Varatonal analyss, Sprnger Verlag, Berln, [41] Solodov, M. V. and Svater, B. F.: Forcng strong convergence of proxmal pont teratons n a Hlbert space, Math. Program , [42] We, L. and Zhou, H. Y.: Projecton scheme for zero ponts of maxmal monotone operators n Banach spaces, J. Math. Res. Exposton , [43] Yao, J. C. and Zeng, L. C.: An nexact proxmal-type algorthm n Banach spaces, J. Optm. Theory Appl , [44] Zălnescu, C.: Convex analyss n general vector spaces, World Scentfc Publshng, Sngapore, Smeon Rech: Department of Mathematcs, The Technon - Israel Insttute of Technology, Hafa, Israel E-mal address: srech@tx.technon.ac.l Shoham Sabach: Department of Mathematcs, The Technon - Israel Insttute of Technology, Hafa, Israel E-mal address: ssabach@tx.technon.ac.l

General viscosity iterative method for a sequence of quasi-nonexpansive mappings

General viscosity iterative method for a sequence of quasi-nonexpansive mappings Avalable onlne at www.tjnsa.com J. Nonlnear Sc. Appl. 9 (2016), 5672 5682 Research Artcle General vscosty teratve method for a sequence of quas-nonexpansve mappngs Cuje Zhang, Ynan Wang College of Scence,

More information

System of implicit nonconvex variationl inequality problems: A projection method approach

System of implicit nonconvex variationl inequality problems: A projection method approach Avalable onlne at www.tjnsa.com J. Nonlnear Sc. Appl. 6 (203), 70 80 Research Artcle System of mplct nonconvex varatonl nequalty problems: A projecton method approach K.R. Kazm a,, N. Ahmad b, S.H. Rzv

More information

More metrics on cartesian products

More metrics on cartesian products More metrcs on cartesan products If (X, d ) are metrc spaces for 1 n, then n Secton II4 of the lecture notes we defned three metrcs on X whose underlyng topologes are the product topology The purpose of

More information

STEINHAUS PROPERTY IN BANACH LATTICES

STEINHAUS PROPERTY IN BANACH LATTICES DEPARTMENT OF MATHEMATICS TECHNICAL REPORT STEINHAUS PROPERTY IN BANACH LATTICES DAMIAN KUBIAK AND DAVID TIDWELL SPRING 2015 No. 2015-1 TENNESSEE TECHNOLOGICAL UNIVERSITY Cookevlle, TN 38505 STEINHAUS

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 ) Kangweon-Kyungk Math. Jour. 4 1996), No. 1, pp. 7 16 AN ITERATIVE ROW-ACTION METHOD FOR MULTICOMMODITY TRANSPORTATION PROBLEMS Yong Joon Ryang Abstract. The optmzaton problems wth quadratc constrants often

More information

ON A DETERMINATION OF THE INITIAL FUNCTIONS FROM THE OBSERVED VALUES OF THE BOUNDARY FUNCTIONS FOR THE SECOND-ORDER HYPERBOLIC EQUATION

ON A DETERMINATION OF THE INITIAL FUNCTIONS FROM THE OBSERVED VALUES OF THE BOUNDARY FUNCTIONS FOR THE SECOND-ORDER HYPERBOLIC EQUATION Advanced Mathematcal Models & Applcatons Vol.3, No.3, 2018, pp.215-222 ON A DETERMINATION OF THE INITIAL FUNCTIONS FROM THE OBSERVED VALUES OF THE BOUNDARY FUNCTIONS FOR THE SECOND-ORDER HYPERBOLIC EUATION

More information

FACTORIZATION IN KRULL MONOIDS WITH INFINITE CLASS GROUP

FACTORIZATION IN KRULL MONOIDS WITH INFINITE CLASS GROUP C O L L O Q U I U M M A T H E M A T I C U M VOL. 80 1999 NO. 1 FACTORIZATION IN KRULL MONOIDS WITH INFINITE CLASS GROUP BY FLORIAN K A I N R A T H (GRAZ) Abstract. Let H be a Krull monod wth nfnte class

More information

APPENDIX A Some Linear Algebra

APPENDIX A Some Linear Algebra APPENDIX A Some Lnear Algebra The collecton of m, n matrces A.1 Matrces a 1,1,..., a 1,n A = a m,1,..., a m,n wth real elements a,j s denoted by R m,n. If n = 1 then A s called a column vector. Smlarly,

More information

Deriving the X-Z Identity from Auxiliary Space Method

Deriving the X-Z Identity from Auxiliary Space Method Dervng the X-Z Identty from Auxlary Space Method Long Chen Department of Mathematcs, Unversty of Calforna at Irvne, Irvne, CA 92697 chenlong@math.uc.edu 1 Iteratve Methods In ths paper we dscuss teratve

More information

Maximizing the number of nonnegative subsets

Maximizing the number of nonnegative subsets Maxmzng the number of nonnegatve subsets Noga Alon Hao Huang December 1, 213 Abstract Gven a set of n real numbers, f the sum of elements of every subset of sze larger than k s negatve, what s the maxmum

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

REAL ANALYSIS I HOMEWORK 1

REAL ANALYSIS I HOMEWORK 1 REAL ANALYSIS I HOMEWORK CİHAN BAHRAN The questons are from Tao s text. Exercse 0.0.. If (x α ) α A s a collecton of numbers x α [0, + ] such that x α

More information

The proximal average for saddle functions and its symmetry properties with respect to partial and saddle conjugacy

The proximal average for saddle functions and its symmetry properties with respect to partial and saddle conjugacy The proxmal average for saddle functons and ts symmetry propertes wth respect to partal and saddle conjugacy Rafal Goebel December 3, 2009 Abstract The concept of the proxmal average for convex functons

More information

Geometry of Müntz Spaces

Geometry of Müntz Spaces WDS'12 Proceedngs of Contrbuted Papers, Part I, 31 35, 212. ISBN 978-8-7378-224-5 MATFYZPRESS Geometry of Müntz Spaces P. Petráček Charles Unversty, Faculty of Mathematcs and Physcs, Prague, Czech Republc.

More information

MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS

MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS These are nformal notes whch cover some of the materal whch s not n the course book. The man purpose s to gve a number of nontrval examples

More information

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg prnceton unv. F 17 cos 521: Advanced Algorthm Desgn Lecture 7: LP Dualty Lecturer: Matt Wenberg Scrbe: LP Dualty s an extremely useful tool for analyzng structural propertes of lnear programs. Whle there

More information

Convexity preserving interpolation by splines of arbitrary degree

Convexity preserving interpolation by splines of arbitrary degree Computer Scence Journal of Moldova, vol.18, no.1(52), 2010 Convexty preservng nterpolaton by splnes of arbtrary degree Igor Verlan Abstract In the present paper an algorthm of C 2 nterpolaton of dscrete

More information

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space.

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space. Lnear, affne, and convex sets and hulls In the sequel, unless otherwse specfed, X wll denote a real vector space. Lnes and segments. Gven two ponts x, y X, we defne xy = {x + t(y x) : t R} = {(1 t)x +

More information

Perfect Competition and the Nash Bargaining Solution

Perfect Competition and the Nash Bargaining Solution Perfect Competton and the Nash Barganng Soluton Renhard John Department of Economcs Unversty of Bonn Adenauerallee 24-42 53113 Bonn, Germany emal: rohn@un-bonn.de May 2005 Abstract For a lnear exchange

More information

A CHARACTERIZATION OF ADDITIVE DERIVATIONS ON VON NEUMANN ALGEBRAS

A CHARACTERIZATION OF ADDITIVE DERIVATIONS ON VON NEUMANN ALGEBRAS Journal of Mathematcal Scences: Advances and Applcatons Volume 25, 2014, Pages 1-12 A CHARACTERIZATION OF ADDITIVE DERIVATIONS ON VON NEUMANN ALGEBRAS JIA JI, WEN ZHANG and XIAOFEI QI Department of Mathematcs

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Another converse of Jensen s inequality

Another converse of Jensen s inequality Another converse of Jensen s nequalty Slavko Smc Abstract. We gve the best possble global bounds for a form of dscrete Jensen s nequalty. By some examples ts frutfulness s shown. 1. Introducton Throughout

More information

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0 MODULE 2 Topcs: Lnear ndependence, bass and dmenson We have seen that f n a set of vectors one vector s a lnear combnaton of the remanng vectors n the set then the span of the set s unchanged f that vector

More information

A note on almost sure behavior of randomly weighted sums of φ-mixing random variables with φ-mixing weights

A note on almost sure behavior of randomly weighted sums of φ-mixing random variables with φ-mixing weights ACTA ET COMMENTATIONES UNIVERSITATIS TARTUENSIS DE MATHEMATICA Volume 7, Number 2, December 203 Avalable onlne at http://acutm.math.ut.ee A note on almost sure behavor of randomly weghted sums of φ-mxng

More information

Feature Selection: Part 1

Feature Selection: Part 1 CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?

More information

Perron Vectors of an Irreducible Nonnegative Interval Matrix

Perron Vectors of an Irreducible Nonnegative Interval Matrix Perron Vectors of an Irreducble Nonnegatve Interval Matrx Jr Rohn August 4 2005 Abstract As s well known an rreducble nonnegatve matrx possesses a unquely determned Perron vector. As the man result of

More information

The Minimum Universal Cost Flow in an Infeasible Flow Network

The Minimum Universal Cost Flow in an Infeasible Flow Network Journal of Scences, Islamc Republc of Iran 17(2): 175-180 (2006) Unversty of Tehran, ISSN 1016-1104 http://jscencesutacr The Mnmum Unversal Cost Flow n an Infeasble Flow Network H Saleh Fathabad * M Bagheran

More information

Duality and Auxiliary Functions for Bregman Distances

Duality and Auxiliary Functions for Bregman Distances Dualty and Auxlary Functons for Bregman Dstances Stephen Della Petra, Vncent Della Petra, and John Lafferty October 8, 2001 CMU-CS-01-109 School of Computer Scence Carnege Mellon Unversty Pttsburgh, PA

More information

ON THE EXTENDED HAAGERUP TENSOR PRODUCT IN OPERATOR SPACES. 1. Introduction

ON THE EXTENDED HAAGERUP TENSOR PRODUCT IN OPERATOR SPACES. 1. Introduction ON THE EXTENDED HAAGERUP TENSOR PRODUCT IN OPERATOR SPACES TAKASHI ITOH AND MASARU NAGISA Abstract We descrbe the Haagerup tensor product l h l and the extended Haagerup tensor product l eh l n terms of

More information

Exercise Solutions to Real Analysis

Exercise Solutions to Real Analysis xercse Solutons to Real Analyss Note: References refer to H. L. Royden, Real Analyss xersze 1. Gven any set A any ɛ > 0, there s an open set O such that A O m O m A + ɛ. Soluton 1. If m A =, then there

More information

MAT 578 Functional Analysis

MAT 578 Functional Analysis MAT 578 Functonal Analyss John Qugg Fall 2008 Locally convex spaces revsed September 6, 2008 Ths secton establshes the fundamental propertes of locally convex spaces. Acknowledgment: although I wrote these

More information

On the Global Linear Convergence of the ADMM with Multi-Block Variables

On the Global Linear Convergence of the ADMM with Multi-Block Variables On the Global Lnear Convergence of the ADMM wth Mult-Block Varables Tany Ln Shqan Ma Shuzhong Zhang May 31, 01 Abstract The alternatng drecton method of multplers ADMM has been wdely used for solvng structured

More information

Dirichlet s Theorem In Arithmetic Progressions

Dirichlet s Theorem In Arithmetic Progressions Drchlet s Theorem In Arthmetc Progressons Parsa Kavkan Hang Wang The Unversty of Adelade February 26, 205 Abstract The am of ths paper s to ntroduce and prove Drchlet s theorem n arthmetc progressons,

More information

Affine transformations and convexity

Affine transformations and convexity Affne transformatons and convexty The purpose of ths document s to prove some basc propertes of affne transformatons nvolvng convex sets. Here are a few onlne references for background nformaton: http://math.ucr.edu/

More information

A Hybrid Variational Iteration Method for Blasius Equation

A Hybrid Variational Iteration Method for Blasius Equation Avalable at http://pvamu.edu/aam Appl. Appl. Math. ISSN: 1932-9466 Vol. 10, Issue 1 (June 2015), pp. 223-229 Applcatons and Appled Mathematcs: An Internatonal Journal (AAM) A Hybrd Varatonal Iteraton Method

More information

Infinitely Split Nash Equilibrium Problems in Repeated Games

Infinitely Split Nash Equilibrium Problems in Repeated Games Infntely Splt ash Equlbrum Problems n Repeated Games Jnlu L Department of Mathematcs Shawnee State Unversty Portsmouth, Oho 4566 USA Abstract In ths paper, we ntroduce the concept of nfntely splt ash equlbrum

More information

SELECTED SOLUTIONS, SECTION (Weak duality) Prove that the primal and dual values p and d defined by equations (4.3.2) and (4.3.3) satisfy p d.

SELECTED SOLUTIONS, SECTION (Weak duality) Prove that the primal and dual values p and d defined by equations (4.3.2) and (4.3.3) satisfy p d. SELECTED SOLUTIONS, SECTION 4.3 1. Weak dualty Prove that the prmal and dual values p and d defned by equatons 4.3. and 4.3.3 satsfy p d. We consder an optmzaton problem of the form The Lagrangan for ths

More information

Genericity of Critical Types

Genericity of Critical Types Genercty of Crtcal Types Y-Chun Chen Alfredo D Tllo Eduardo Fangold Syang Xong September 2008 Abstract Ely and Pesk 2008 offers an nsghtful characterzaton of crtcal types: a type s crtcal f and only f

More information

Appendix B. Criterion of Riemann-Stieltjes Integrability

Appendix B. Criterion of Riemann-Stieltjes Integrability Appendx B. Crteron of Remann-Steltes Integrablty Ths note s complementary to [R, Ch. 6] and [T, Sec. 3.5]. The man result of ths note s Theorem B.3, whch provdes the necessary and suffcent condtons for

More information

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009 College of Computer & Informaton Scence Fall 2009 Northeastern Unversty 20 October 2009 CS7880: Algorthmc Power Tools Scrbe: Jan Wen and Laura Poplawsk Lecture Outlne: Prmal-dual schema Network Desgn:

More information

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS BOUNDEDNESS OF THE IESZ TANSFOM WITH MATIX A WEIGHTS Introducton Let L = L ( n, be the functon space wth norm (ˆ f L = f(x C dx d < For a d d matrx valued functon W : wth W (x postve sem-defnte for all

More information

On C 0 multi-contractions having a regular dilation

On C 0 multi-contractions having a regular dilation SUDIA MAHEMAICA 170 (3) (2005) On C 0 mult-contractons havng a regular dlaton by Dan Popovc (mşoara) Abstract. Commutng mult-contractons of class C 0 and havng a regular sometrc dlaton are studed. We prove

More information

Self-complementing permutations of k-uniform hypergraphs

Self-complementing permutations of k-uniform hypergraphs Dscrete Mathematcs Theoretcal Computer Scence DMTCS vol. 11:1, 2009, 117 124 Self-complementng permutatons of k-unform hypergraphs Artur Szymańsk A. Paweł Wojda Faculty of Appled Mathematcs, AGH Unversty

More information

Research Article Relative Smooth Topological Spaces

Research Article Relative Smooth Topological Spaces Advances n Fuzzy Systems Volume 2009, Artcle ID 172917, 5 pages do:10.1155/2009/172917 Research Artcle Relatve Smooth Topologcal Spaces B. Ghazanfar Department of Mathematcs, Faculty of Scence, Lorestan

More information

10-801: Advanced Optimization and Randomized Methods Lecture 2: Convex functions (Jan 15, 2014)

10-801: Advanced Optimization and Randomized Methods Lecture 2: Convex functions (Jan 15, 2014) 0-80: Advanced Optmzaton and Randomzed Methods Lecture : Convex functons (Jan 5, 04) Lecturer: Suvrt Sra Addr: Carnege Mellon Unversty, Sprng 04 Scrbes: Avnava Dubey, Ahmed Hefny Dsclamer: These notes

More information

Randić Energy and Randić Estrada Index of a Graph

Randić Energy and Randić Estrada Index of a Graph EUROPEAN JOURNAL OF PURE AND APPLIED MATHEMATICS Vol. 5, No., 202, 88-96 ISSN 307-5543 www.ejpam.com SPECIAL ISSUE FOR THE INTERNATIONAL CONFERENCE ON APPLIED ANALYSIS AND ALGEBRA 29 JUNE -02JULY 20, ISTANBUL

More information

Uniqueness of Weak Solutions to the 3D Ginzburg- Landau Model for Superconductivity

Uniqueness of Weak Solutions to the 3D Ginzburg- Landau Model for Superconductivity Int. Journal of Math. Analyss, Vol. 6, 212, no. 22, 195-114 Unqueness of Weak Solutons to the 3D Gnzburg- Landau Model for Superconductvty Jshan Fan Department of Appled Mathematcs Nanjng Forestry Unversty

More information

A note on the Legendre-Fenchel transform of convex composite functions.

A note on the Legendre-Fenchel transform of convex composite functions. A note on the Legendre-Fenchel transform of convex composte functons. J.-B. Hrart-Urruty Unversté Paul Sabater 118, route de Narbonne 31062 Toulouse cedex 4, France. jbhu@cct.fr Dedcated to J.-J. MOREAU

More information

DIEGO AVERNA. A point x 2 X is said to be weakly Pareto-optimal for the function f provided

DIEGO AVERNA. A point x 2 X is said to be weakly Pareto-optimal for the function f provided WEAKLY PARETO-OPTIMAL ALTERNATIVES FOR A VECTOR MAXIMIZATION PROBLEM: EXISTENCE AND CONNECTEDNESS DIEGO AVERNA Let X be a non-empty set and f =f 1 ::: f k ):X! R k a functon. A pont x 2 X s sad to be weakly

More information

Strong Convergence Theorems for a Common Fixed Point of a Finite Family of Multivalued Mappings in Certain Banach Spaces

Strong Convergence Theorems for a Common Fixed Point of a Finite Family of Multivalued Mappings in Certain Banach Spaces Internatonal Journal of Mathematcal Analyss Vol. 9, 2015, no. 9, 437-452 HIKARI Ltd, www.m-hkar.com http://dx.do.org/10.12988/jma.2015.3386 Strong Convergence Theorems for a Common Fxed Pont of a Fnte

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

Generalized Linear Methods

Generalized Linear Methods Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set

More information

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b Int J Contemp Math Scences, Vol 3, 28, no 17, 819-827 A New Refnement of Jacob Method for Soluton of Lnear System Equatons AX=b F Naem Dafchah Department of Mathematcs, Faculty of Scences Unversty of Gulan,

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso

Supplement: Proofs and Technical Details for The Solution Path of the Generalized Lasso Supplement: Proofs and Techncal Detals for The Soluton Path of the Generalzed Lasso Ryan J. Tbshran Jonathan Taylor In ths document we gve supplementary detals to the paper The Soluton Path of the Generalzed

More information

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix Lectures - Week 4 Matrx norms, Condtonng, Vector Spaces, Lnear Independence, Spannng sets and Bass, Null space and Range of a Matrx Matrx Norms Now we turn to assocatng a number to each matrx. We could

More information

P.P. PROPERTIES OF GROUP RINGS. Libo Zan and Jianlong Chen

P.P. PROPERTIES OF GROUP RINGS. Libo Zan and Jianlong Chen Internatonal Electronc Journal of Algebra Volume 3 2008 7-24 P.P. PROPERTIES OF GROUP RINGS Lbo Zan and Janlong Chen Receved: May 2007; Revsed: 24 October 2007 Communcated by John Clark Abstract. A rng

More information

SUCCESSIVE MINIMA AND LATTICE POINTS (AFTER HENK, GILLET AND SOULÉ) M(B) := # ( B Z N)

SUCCESSIVE MINIMA AND LATTICE POINTS (AFTER HENK, GILLET AND SOULÉ) M(B) := # ( B Z N) SUCCESSIVE MINIMA AND LATTICE POINTS (AFTER HENK, GILLET AND SOULÉ) S.BOUCKSOM Abstract. The goal of ths note s to present a remarably smple proof, due to Hen, of a result prevously obtaned by Gllet-Soulé,

More information

The lower and upper bounds on Perron root of nonnegative irreducible matrices

The lower and upper bounds on Perron root of nonnegative irreducible matrices Journal of Computatonal Appled Mathematcs 217 (2008) 259 267 wwwelsevercom/locate/cam The lower upper bounds on Perron root of nonnegatve rreducble matrces Guang-Xn Huang a,, Feng Yn b,keguo a a College

More information

SUPER PRINCIPAL FIBER BUNDLE WITH SUPER ACTION

SUPER PRINCIPAL FIBER BUNDLE WITH SUPER ACTION talan journal of pure appled mathematcs n. 33 2014 (63 70) 63 SUPER PRINCIPAL FIBER BUNDLE WITH SUPER ACTION M.R. Farhangdoost Department of Mathematcs College of Scences Shraz Unversty Shraz, 71457-44776

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

On a direct solver for linear least squares problems

On a direct solver for linear least squares problems ISSN 2066-6594 Ann. Acad. Rom. Sc. Ser. Math. Appl. Vol. 8, No. 2/2016 On a drect solver for lnear least squares problems Constantn Popa Abstract The Null Space (NS) algorthm s a drect solver for lnear

More information

Fixed point method and its improvement for the system of Volterra-Fredholm integral equations of the second kind

Fixed point method and its improvement for the system of Volterra-Fredholm integral equations of the second kind MATEMATIKA, 217, Volume 33, Number 2, 191 26 c Penerbt UTM Press. All rghts reserved Fxed pont method and ts mprovement for the system of Volterra-Fredholm ntegral equatons of the second knd 1 Talaat I.

More information

On the Operation A in Analysis Situs. by Kazimierz Kuratowski

On the Operation A in Analysis Situs. by Kazimierz Kuratowski v1.3 10/17 On the Operaton A n Analyss Stus by Kazmerz Kuratowsk Author s note. Ths paper s the frst part slghtly modfed of my thess presented May 12, 1920 at the Unversty of Warsaw for the degree of Doctor

More information

ON SEPARATING SETS OF WORDS IV

ON SEPARATING SETS OF WORDS IV ON SEPARATING SETS OF WORDS IV V. FLAŠKA, T. KEPKA AND J. KORTELAINEN Abstract. Further propertes of transtve closures of specal replacement relatons n free monods are studed. 1. Introducton Ths artcle

More information

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES BÂRZĂ, Slvu Faculty of Mathematcs-Informatcs Spru Haret Unversty barza_slvu@yahoo.com Abstract Ths paper wants to contnue

More information

THE WEIGHTED WEAK TYPE INEQUALITY FOR THE STRONG MAXIMAL FUNCTION

THE WEIGHTED WEAK TYPE INEQUALITY FOR THE STRONG MAXIMAL FUNCTION THE WEIGHTED WEAK TYPE INEQUALITY FO THE STONG MAXIMAL FUNCTION THEMIS MITSIS Abstract. We prove the natural Fefferman-Sten weak type nequalty for the strong maxmal functon n the plane, under the assumpton

More information

THE STRUCTURED DISTANCE TO ILL-POSEDNESS FOR CONIC SYSTEMS. A.S. Lewis. April 18, 2003

THE STRUCTURED DISTANCE TO ILL-POSEDNESS FOR CONIC SYSTEMS. A.S. Lewis. April 18, 2003 THE STRUCTURED DISTANCE TO ILL-POSEDNESS FOR CONIC SYSTEMS A.S. Lews Aprl 18, 2003 Key words: condton number, conc system, dstance to nfeasblty, structured sngular values, sublnear maps, surjectvty AMS

More information

n ). This is tight for all admissible values of t, k and n. k t + + n t

n ). This is tight for all admissible values of t, k and n. k t + + n t MAXIMIZING THE NUMBER OF NONNEGATIVE SUBSETS NOGA ALON, HAROUT AYDINIAN, AND HAO HUANG Abstract. Gven a set of n real numbers, f the sum of elements of every subset of sze larger than k s negatve, what

More information

Asymptotics of the Solution of a Boundary Value. Problem for One-Characteristic Differential. Equation Degenerating into a Parabolic Equation

Asymptotics of the Solution of a Boundary Value. Problem for One-Characteristic Differential. Equation Degenerating into a Parabolic Equation Nonl. Analyss and Dfferental Equatons, ol., 4, no., 5 - HIKARI Ltd, www.m-har.com http://dx.do.org/.988/nade.4.456 Asymptotcs of the Soluton of a Boundary alue Problem for One-Characterstc Dfferental Equaton

More information

Homogenization of reaction-diffusion processes in a two-component porous medium with a non-linear flux-condition on the interface

Homogenization of reaction-diffusion processes in a two-component porous medium with a non-linear flux-condition on the interface Homogenzaton of reacton-dffuson processes n a two-component porous medum wth a non-lnear flux-condton on the nterface Internatonal Conference on Numercal and Mathematcal Modelng of Flow and Transport n

More information

12 MATH 101A: ALGEBRA I, PART C: MULTILINEAR ALGEBRA. 4. Tensor product

12 MATH 101A: ALGEBRA I, PART C: MULTILINEAR ALGEBRA. 4. Tensor product 12 MATH 101A: ALGEBRA I, PART C: MULTILINEAR ALGEBRA Here s an outlne of what I dd: (1) categorcal defnton (2) constructon (3) lst of basc propertes (4) dstrbutve property (5) rght exactness (6) localzaton

More information

On the Multicriteria Integer Network Flow Problem

On the Multicriteria Integer Network Flow Problem BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 5, No 2 Sofa 2005 On the Multcrtera Integer Network Flow Problem Vassl Vasslev, Marana Nkolova, Maryana Vassleva Insttute of

More information

Lecture 20: Lift and Project, SDP Duality. Today we will study the Lift and Project method. Then we will prove the SDP duality theorem.

Lecture 20: Lift and Project, SDP Duality. Today we will study the Lift and Project method. Then we will prove the SDP duality theorem. prnceton u. sp 02 cos 598B: algorthms and complexty Lecture 20: Lft and Project, SDP Dualty Lecturer: Sanjeev Arora Scrbe:Yury Makarychev Today we wll study the Lft and Project method. Then we wll prove

More information

Y. Guo. A. Liu, T. Liu, Q. Ma UDC

Y. Guo. A. Liu, T. Liu, Q. Ma UDC UDC 517. 9 OSCILLATION OF A CLASS OF NONLINEAR PARTIAL DIFFERENCE EQUATIONS WITH CONTINUOUS VARIABLES* ОСЦИЛЯЦIЯ КЛАСУ НЕЛIНIЙНИХ ЧАСТКОВО РIЗНИЦЕВИХ РIВНЯНЬ З НЕПЕРЕРВНИМИ ЗМIННИМИ Y. Guo Graduate School

More information

Convergence rates of proximal gradient methods via the convex conjugate

Convergence rates of proximal gradient methods via the convex conjugate Convergence rates of proxmal gradent methods va the convex conjugate Davd H Gutman Javer F Peña January 8, 018 Abstract We gve a novel proof of the O(1/ and O(1/ convergence rates of the proxmal gradent

More information

arxiv: v1 [math.oc] 6 Jan 2016

arxiv: v1 [math.oc] 6 Jan 2016 arxv:1601.01174v1 [math.oc] 6 Jan 2016 THE SUPPORTING HALFSPACE - QUADRATIC PROGRAMMING STRATEGY FOR THE DUAL OF THE BEST APPROXIMATION PROBLEM C.H. JEFFREY PANG Abstract. We consder the best approxmaton

More information

arxiv: v1 [math.co] 12 Sep 2014

arxiv: v1 [math.co] 12 Sep 2014 arxv:1409.3707v1 [math.co] 12 Sep 2014 On the bnomal sums of Horadam sequence Nazmye Ylmaz and Necat Taskara Department of Mathematcs, Scence Faculty, Selcuk Unversty, 42075, Campus, Konya, Turkey March

More information

Random Projection Algorithms for Convex Set Intersection Problems

Random Projection Algorithms for Convex Set Intersection Problems Random Projecton Algorthms for Convex Set Intersecton Problems A. Nedć Department of Industral and Enterprse Systems Engneerng Unversty of Illnos, Urbana, IL 61801 angela@llnos.edu Abstract The focus of

More information

arxiv: v1 [math.co] 1 Mar 2014

arxiv: v1 [math.co] 1 Mar 2014 Unon-ntersectng set systems Gyula O.H. Katona and Dánel T. Nagy March 4, 014 arxv:1403.0088v1 [math.co] 1 Mar 014 Abstract Three ntersecton theorems are proved. Frst, we determne the sze of the largest

More information

First day August 1, Problems and Solutions

First day August 1, Problems and Solutions FOURTH INTERNATIONAL COMPETITION FOR UNIVERSITY STUDENTS IN MATHEMATICS July 30 August 4, 997, Plovdv, BULGARIA Frst day August, 997 Problems and Solutons Problem. Let {ε n } n= be a sequence of postve

More information

Some modelling aspects for the Matlab implementation of MMA

Some modelling aspects for the Matlab implementation of MMA Some modellng aspects for the Matlab mplementaton of MMA Krster Svanberg krlle@math.kth.se Optmzaton and Systems Theory Department of Mathematcs KTH, SE 10044 Stockholm September 2004 1. Consdered optmzaton

More information

On Finite Rank Perturbation of Diagonalizable Operators

On Finite Rank Perturbation of Diagonalizable Operators Functonal Analyss, Approxmaton and Computaton 6 (1) (2014), 49 53 Publshed by Faculty of Scences and Mathematcs, Unversty of Nš, Serba Avalable at: http://wwwpmfnacrs/faac On Fnte Rank Perturbaton of Dagonalzable

More information

20. Mon, Oct. 13 What we have done so far corresponds roughly to Chapters 2 & 3 of Lee. Now we turn to Chapter 4. The first idea is connectedness.

20. Mon, Oct. 13 What we have done so far corresponds roughly to Chapters 2 & 3 of Lee. Now we turn to Chapter 4. The first idea is connectedness. 20. Mon, Oct. 13 What we have done so far corresponds roughly to Chapters 2 & 3 of Lee. Now we turn to Chapter 4. The frst dea s connectedness. Essentally, we want to say that a space cannot be decomposed

More information

Introductory Cardinality Theory Alan Kaylor Cline

Introductory Cardinality Theory Alan Kaylor Cline Introductory Cardnalty Theory lan Kaylor Clne lthough by name the theory of set cardnalty may seem to be an offshoot of combnatorcs, the central nterest s actually nfnte sets. Combnatorcs deals wth fnte

More information

Fixed points of IA-endomorphisms of a free metabelian Lie algebra

Fixed points of IA-endomorphisms of a free metabelian Lie algebra Proc. Indan Acad. Sc. (Math. Sc.) Vol. 121, No. 4, November 2011, pp. 405 416. c Indan Academy of Scences Fxed ponts of IA-endomorphsms of a free metabelan Le algebra NAIME EKICI 1 and DEMET PARLAK SÖNMEZ

More information

TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES

TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES SVANTE JANSON Abstract. We gve explct bounds for the tal probabltes for sums of ndependent geometrc or exponental varables, possbly wth dfferent

More information

FINITELY-GENERATED MODULES OVER A PRINCIPAL IDEAL DOMAIN

FINITELY-GENERATED MODULES OVER A PRINCIPAL IDEAL DOMAIN FINITELY-GENERTED MODULES OVER PRINCIPL IDEL DOMIN EMMNUEL KOWLSKI Throughout ths note, s a prncpal deal doman. We recall the classfcaton theorem: Theorem 1. Let M be a fntely-generated -module. (1) There

More information

Christian Aebi Collège Calvin, Geneva, Switzerland

Christian Aebi Collège Calvin, Geneva, Switzerland #A7 INTEGERS 12 (2012) A PROPERTY OF TWIN PRIMES Chrstan Aeb Collège Calvn, Geneva, Swtzerland chrstan.aeb@edu.ge.ch Grant Carns Department of Mathematcs, La Trobe Unversty, Melbourne, Australa G.Carns@latrobe.edu.au

More information

Curvature and isoperimetric inequality

Curvature and isoperimetric inequality urvature and sopermetrc nequalty Julà ufí, Agustí Reventós, arlos J Rodríguez Abstract We prove an nequalty nvolvng the length of a plane curve and the ntegral of ts radus of curvature, that has as a consequence

More information

THE CHVÁTAL-ERDŐS CONDITION AND 2-FACTORS WITH A SPECIFIED NUMBER OF COMPONENTS

THE CHVÁTAL-ERDŐS CONDITION AND 2-FACTORS WITH A SPECIFIED NUMBER OF COMPONENTS Dscussones Mathematcae Graph Theory 27 (2007) 401 407 THE CHVÁTAL-ERDŐS CONDITION AND 2-FACTORS WITH A SPECIFIED NUMBER OF COMPONENTS Guantao Chen Department of Mathematcs and Statstcs Georga State Unversty,

More information

arxiv:quant-ph/ Feb 2000

arxiv:quant-ph/ Feb 2000 Entanglement measures and the Hlbert-Schmdt dstance Masanao Ozawa School of Informatcs and Scences, Nagoya Unversty, Chkusa-ku, Nagoya 464-86, Japan Abstract arxv:quant-ph/236 3 Feb 2 In order to construct

More information

On the Connectedness of the Solution Set for the Weak Vector Variational Inequality 1

On the Connectedness of the Solution Set for the Weak Vector Variational Inequality 1 Journal of Mathematcal Analyss and Alcatons 260, 15 2001 do:10.1006jmaa.2000.7389, avalable onlne at htt:.dealbrary.com on On the Connectedness of the Soluton Set for the Weak Vector Varatonal Inequalty

More information

A Douglas-Rachford type primal-dual method for solving inclusions with mixtures of composite and parallel-sum type monotone operators

A Douglas-Rachford type primal-dual method for solving inclusions with mixtures of composite and parallel-sum type monotone operators A Douglas-Rachford type prmal-dual method for solvng nclusons wth mxtures of composte and parallel-sum type monotone operators Radu Ioan Boţ Chrstopher Hendrch October 2, 2013 Abstract. In ths paper we

More information

Existence results for a fourth order multipoint boundary value problem at resonance

Existence results for a fourth order multipoint boundary value problem at resonance Avalable onlne at www.scencedrect.com ScenceDrect Journal of the Ngeran Mathematcal Socety xx (xxxx) xxx xxx www.elsever.com/locate/jnnms Exstence results for a fourth order multpont boundary value problem

More information

a b a In case b 0, a being divisible by b is the same as to say that

a b a In case b 0, a being divisible by b is the same as to say that Secton 6.2 Dvsblty among the ntegers An nteger a ε s dvsble by b ε f there s an nteger c ε such that a = bc. Note that s dvsble by any nteger b, snce = b. On the other hand, a s dvsble by only f a = :

More information