ASYMPTOTIC BEHAVIOR OF CONTINUOUS TRAJECTORIES FOR PRIMAL-DUAL POTENTIAL-REDUCTION METHODS
|
|
- Kelly Dean
- 6 years ago
- Views:
Transcription
1 ASYMPTOTIC EHAVIOR OF CONTINUOUS TRAJECTORIES FOR PRIMAL-DUAL POTENTIAL-REDUCTION METHODS REHA H. TÜTÜNCÜ Abstract. This article cosiders cotiuous trajectories of the vector fields iduced by primaldual potetial-reductio algorithms for solvig liear programmig problems. It is kow that these trajectories coverge to the aalytic ceter of the primal-dual optimal face. We establish that this covergece may be tagetial to the cetral path, tagetial to the optimal face, or i betwee, depedig o the value of the potetial fuctio parameter. Key words. liear programmig, potetial fuctios, potetial-reductio methods, cetral path, cotiuous trajectories for liear programmig. AMS subject classificatios. 90C05 1. Itroductio. Durig the past two decades, iterior-poit methods IPMs emerged as oe of the most efficiet ad reliable techiques for the solutio of liear programmig problems. The developmet of IPMs ad their theoretical covergece aalyses ofte rely o certai cotiuous trajectories associated with the give liear program. The best kow examples of such trajectories are the cetral path ad the weighted ceters the sets of miimizers of the parametrized stadard ad weighted logarithmic barrier fuctios i the iterior of the feasible regio. Primal-dual variats of IPMs, which have bee very successful i practical implemetatios, ot oly solve the give liear program but also its dual. If both the give LP ad its dual have strictly feasible solutios the primal-dual cetral path starts from the aalytic ceter of the primal-dual feasible set ad coverges to the aalytic ceter of the optimal solutio set. Similarly, weighted ceters coverge to weighted aalytic ceters. This property of the cetral trajectories led to the developmet of path followig IPMs: algorithms that try to reach a optimal solutio by geeratig a sequece of poits that are close to a correspodig sequece of poits o the cetral path or the weighted cetral path that coverge to its limit poit. A alterative characterizatio of the cetral path ad weighted ceters ca be obtaied by represetig them as solutios of certai differetial equatios. Usig this perspective, Adler ad Moteiro aalyzed the limitig behavior of cotiuous trajectories associated with primal-oly affie-scalig ad projective-scalig algorithms as well as a primal-oly potetial-reductio method [1, 11, 12]. Kojima et al. studied similar trajectories for primal-dual potetial-reductio methods [7]. Potetial-reductio algorithms use the followig strategy: First, oe defies a potetial fuctio that measures the quality or potetial of ay trial solutio of the give problem combiig measures of proximity to the set of optimal solutios, proximity to the feasible set i the case of ifeasible-iterior-poits, ad a measure of cetrality withi the feasible regio. Potetial fuctios are chose such that oe approaches a optimal solutio of the uderlyig problem by reducig the potetial fuctio. The, the search for a optimal solutio ca be performed by progressive reductio of the potetial fuctio, leadig to a potetial-reductio algorithm. We refer the reader to two excellet surveys for further details o potetial-reductio algorithms [2, 15]. Departmet of Mathematical Scieces, Caregie Mello Uiversity, Pittsburgh, PA 15213, USA reha@cmu.edu. Research supported i part by NSF through grat CCR
2 2 REHA H. TÜTÜNCÜ Ofte, implemetatios of potetial-reductio iterior-poit algorithms exhibit behavior that is similar to that of path-followig algorithms. For example, they take about the same umber of iteratios as path-followig algorithms ad they ted to coverge to the aalytic ceter of the optimal face, just like most path-followig variats. Sice potetial-reductio methods do ot geerally make a effort to follow the cetral path, this behavior is surprisig. I a effort to better uderstad the limitig behavior of primal-dual potetial-reductio algorithms for liear programs this paper studies cotiuous trajectories associated with the algorithm proposed by Kojima, Mizuo, ad Yoshise KMY [8], which uses scaled ad projected steepest descet directios for the Taabe-Todd-Ye TTY primal-dual potetial fuctio [14, 16]. Usig earlier results [9, 10, 7], we show that all trajectories of the vector field iduced by the KMY search directios coverge to the aalytic ceter of the primaldual optimal face. Our mai results are o the directio of covergece for these trajectories. We demostrate that their asymptotic behavior depeds o the potetial fuctio parameter. There is a threshold value of this parameter the value that makes the TTY potetial fuctio homogeeous. Whe the parameter is below this threshold, the ceterig is too strog ad the trajectories coverge tagetially to the cetral path. Whe the parameter is above the threshold, trajectories coverge tagetially to the optimal face. However, the directio of covergece of these trajectories depeds o the iitial poit. At the threshold value, the behavior of the trajectories is i betwee these two extremes ad depeds o the iitial poit. Followig this itroductio, Sectio 2 discusses cotiuous trajectories associated with the KMY methods ad proves their covergece. Sectio 3 is devoted to the aalysis of the limitig behavior of these trajectories. Our otatio is fairly stadard: For a -dimesioal vector x, the correspodig capital letter X deotes the diagoal matrix with X ii x i. We will use the letter e to deote a colum vector with all etries equal to 1 ad its dimesio will be apparet from the cotext. We also deote the base of the atural logarithm with e ad sometimes the vector e ad the scalar e appear i the same expressio, but o cofusio should arise. For a give matrix A, we use RA ad N A to deote its ragecolum ad ull space. For a vector-valued differetiable fuctio xt of a scalar variable t, we use the otatio ẋ or ẋt to deote the vector of the derivatives of its compoets with respect to t. For dimesioal vectors x ad s, we write xs to deote their Hadamard compoetwise product. Also, for a dimesioal vector x, we write x p to deote the vector X p e, where p ca be fractioal if x > Primal-Dual Potetial-Reductio Trajectories. We cosider liear programs i the followig stadard form: 2.1 LP mi x c T x Ax = b x 0, where A R m, b R m, c R are give, ad x R. Without loss of geerality we assume that the costraits are liearly idepedet. The, the matrix A has full row rak. Further, we assume that 0 < m < ; m = 0 ad m = correspod to trivial problems.
3 POTENTIAL REDUCTION TRAJECTORIES 3 The liear programmig dual of this primal problem is: 2.2 LD max y,s b T y A T y + s = c s 0, where y R m ad s R. We ca rewrite the dual problem by elimiatig the y variables i 2.2. This is achieved by cosiderig G T, a ull-space basis matrix for A, that is, G is a m matrix with rak m ad it satisfies AG T = 0, GA T = 0. Note also that, A T is a ull-space basis matrix for G. Further, let d R be a vector satisfyig Ad = b. The, 2.2 is equivalet to the followig problem which has a high degree of symmetry with 2.1: 2.3 LD mi s d T s Gs = Gc s 0. Let F ad F 0 deote the primal-dual feasible regio ad its relative iterior: F := {x, s : Ax = b, Gs = Gc, x, s 0}, F 0 := {x, s : Ax = b, Gs = Gc, x, s > 0}. We assume that F 0 is o-empty. This assumptio has the importat cosequece that the primal-dual optimal solutio set Ω defied below is oempty ad bouded: 2.4 Ω := {x, s F : x T s = 0}. We also defie the optimal partitio N = {1,..., } for future referece: := {j : x j > 0 for some x, s Ω}, N := {j : s j > 0 for some x, s Ω}. The fact that ad N is a partitio of {1,..., } is a classical result of Goldma ad Tucker. The aalytic ceter of Ω is the poit x, s = x, 0, 0, s N where x ad s N are uique maximizers of the followig problems: max j l x j max j N l s j 2.5 A x = b ad G N s N = Gc x > 0, s N > 0. The cetral path C of the primal-dual feasible set F is the set of poits o which the compoet-wise product of the primal ad dual variables is costat: 2.6 C := {xµ, sµ F 0 : xµsµ = µe, for some µ > 0}. The poits o the cetral path are obtaied as uique miimizers of certai barrier problems associated with the primal ad dual LPs ad they coverge to the aalytic ceter of the primal-dual optimal face; see, e.g., [18]. While the cetral path is the mai theoretical tool i the costructio of pathfollowig algorithms, primal-dual potetial-reductio algorithms for liear programmig are derived usig potetial fuctios, i.e., fuctios that measure the quality
4 4 REHA H. TÜTÜNCÜ or potetial of trial solutios for the primal-dual pair of problems. The most frequetly used primal-dual potetial fuctio for liear programmig problems is the Taabe-Todd-Ye TTY potetial fuctio [14, 16]: 2.7 Φ ρ x, s := ρ lx T s lx i s i, for every x, s > 0. i=1 Whe ρ >, the TTY potetial fuctio diverges to alog a feasible sequece {x k, s k } oly if this sequece is covergig to a primal-dual optimal pair of solutios. Therefore, the primal-dual pair of LP problems ca be solved by miimizig the TTY potetial fuctio. Kojima, Mizuo, ad Yoshise KMY developed a primal-dual algorithm that mootoically reduces the TTY potetial fuctio usig a scaled ad projected steepestdescet search directio usig a primal-dual scalig matrix [8]. I the remaider of this article, we will study cotiuous trajectories that are aturally associated with their algorithm. Give a iterate x, s F 0, the search directio used by the KMY method is the solutio of the followig system: 2.8 A x = 0 G s = 0 S x + X s = xt s ρ e xs, where X = diagx, S = diags, ad e is a vector of oes of appropriate dimesio. Whe we discuss the search directio give by 2.8 ad associated trajectories, we will assume that ρ >. For ay give x 0, s 0 F 0, oe ca associate a trajectory {xt, st : t 0} startig from x 0, s 0 with the property that the taget directio to the trajectory at ay of its poits coicides with the KMY directio. I other words, we cosider trajectories that solve the followig system of ordiary differetial equatios: 2.9 A 0 [ ] 0 G ẋ = ṡ S X 0 0 x T s ρ e xs with the iitial coditio x0, s0 = x 0, s 0. I [7, Sectio 4.3], Kojima et al. study these trajectories ad establish that their solutio curves satisfy the followig system of equatios:, 2.10 Axt = b, Gst = Gc, xtst = wt, t 0, where wt = e t w 0 + hte, with w 0 = x 0 s 0, ad, ht = et w 0 exp { 1 β t} e t with β = ρ. Sice w0 = w 0, we will use these two expressios iterchageably. Kojima et al. do ot address the existece ad uiqueess of the solutios to 2.9 rigorously, but these results follow easily from stadard theory of ordiary differetial equatios; see,
5 POTENTIAL REDUCTION TRAJECTORIES 5 e.g., Theorem 1 o p. 162 ad Lemma o p. 171 of the textbook by Hirsch ad Smale [6]. We also ote that Moteiro [12] studies trajectories based o primal-oly potetial-reductio algorithms ad obtais similar but less explicit descriptios of these trajectories. The characterizatio of the potetial-reductio trajectories usig the system 2.10 leads to the followig observatios: Theorem 2.1. Let xt, st for t 0 deote the solutio of the ODE 2.9 with the iitial coditio x0, s0 = x 0, s 0. The, the followig statemets hold: i For ρ >, Φ ρ xt, st is a decreasig fuctio of t. ii Whe w 0 = x 0 s 0 = µe for some µ > 0 i.e., whe x 0, s 0 is o the cetral path the {xt, st : t 0} is a subset of the cetral path C. iii xt, st coverges to the aalytic ceter of the primal-dual optimal face Ω as t. Proof. Lemma 4.14 i [7] proves i. Observig that w 0 = µe implies wt = µe 1 βt e with β = ρ, ii follows immediately from 2.6 ad For iii, first wt observe that wt e as t. Now, the proof of Theorem 9 i [9] or, Corollary 2 of Theorem 5 i [10] immediately leads to iii. So, these trajectories coverge to a uique poit regardless of their startig poit as they mootoically decrease the potetial fuctio. Further, they iclude the cetral path as a special case givig a theoretical basis for the observatio that cetral path-followig search directios are ofte very good potetial-reductio directios as well. Aother related result is by Nesterov [13] who observes that the eighborhood of the cetral path is the regio of fastest decrease for a homogeeous potetial fuctio. A direct proof of iii i Theorem 2.1 ca be obtaied usig the followig result, which will also be useful i the ext sectio: Lemma 2.2. Let xt, st for t 0 deote the solutio of the ODE 2.9 with the iitial coditio x0, s0 = x 0, s 0. The x t ad s N t solve the followig pair of problems: 2.13 max j w jt l x j A x = b A N x N t x > 0, ad max j N w jt l s j G N s N = Gc G s t s N > 0. Proof. We prove the optimality of x t for the first problem i 2.13 the correspodig result for s N t ca be prove similarly. x t is clearly feasible for the give problem. It is optimal if ad oly if there exists y R m such that w tx 1 t = AT y. From 2.10 we obtai w tx 1 t = s t. Note that, for ay s feasible for LD we have that c s RA T ad therefore, c s RA T. Furthermore, sice s = 0, s N is also feasible for LD we must have that c RA T ad that s t RA T. This is exactly what we eeded. 3. Asymptotic Aalysis of the Trajectories. I the previous sectio, we saw that all primal-dual potetial-reductio trajectories xt, st that solve the differetial equatio 2.9 coverge to the aalytic ceter x, s of the primal-dual optimal face Ω regardless of the iitial poit of the trajectory. I this sectio, we ivestigate the directio of covergece for these trajectories. That is, we wat to aalyze the limitig behavior of the ormalized vectors this aalysis is quite techical. ẋt ẋt, ṡt ṡt. Ievitably,
6 6 REHA H. TÜTÜNCÜ Our strategy for this aalysis is as follows. Usig the optimal partitio ad N we express the basic compoets of the covergece directios of the trajectories i terms of the obasic oes i Lemma 3.1. The, we establish a boud o the covergece speed of the obasic compoets i Lemma 3.3. The depedece of the covergece directio o ρ, the potetial fuctio parameter, becomes apparet at this poit ad a case aalysis is required. Lemma 3.4, Theorems 3.5 ad 3.6 cosider the case ρ 2 while Theorem 3.8 addresses the case ρ > 2. Let β = ρ ad ote that β 0, 1. We ow itroduce some otatio: ŵ t = w te 1 βt, ŵ N t = w N te 1 βt, d t = ŵ 1 2 tx 1 t, d N t = ŵ 1 2 N ts 1 N t, D t = diag d t, D N t = diag d N t, d 1 t = D 1 te, d 1 N t = D 1 N te, Ã t = A D 1 t, GN t = G N D 1 N t, x t = D tẋ t = d tẋ t, s N t = D N tṡ N t = d N tṡ N t, u t = ŵ 1 2 tw 0, u N t = ŵ 1 2 N tw N 0. For our asymptotic aalysis, we express basic compoets of the vectors ẋt ad ṡt i terms of the obasic oes i the ext lemma which forms the backboe of our aalysis. Lemma 3.1. Let xt, st for t 0 deote the solutio of the ODE 2.9 with the iitial coditio x0, s0 = x 0, s 0. The, the followig equalities hold: D tẋ t = Ã+ ta N ẋ N t e βt ρ 1 e βt D N tṡ N t = G + N tg ṡ t e βt ρ 1 e βt I Ã+ tãt u t, I G + N t G N t u N t. Here, Ã+ t ad G + N t deote the pseudo-iverse of Ãt ad G N t, respectively. Proof. We oly prove the first idetity; the secod oe follows similarly. Recall from Lemma 2.2 that x t solves the first problem i Therefore, as i the proof of Lemma 2.2, we must have that 3.3 w tx 1 t RAT. Differetiatig with respect to t we obtai: Observe that w tx 2 w tx 1 tẋ t ẇ tx 1 t RAT, or, tẋ t ẇ t RX ta T. ẇ t = w 0e t + ḣte = w t + et w 0 ρ e 1 βt e. Therefore, from ŵ t = w te 1 βt, we obtai 3.4 ŵ tx 1 tẋ t + ŵ t et w 0 ρ e RX ta T.
7 POTENTIAL REDUCTION TRAJECTORIES 7 From 3.3 it also follows that ŵ t RX ta T. Note also that, e T w 0 ρ e = ρ 1 e βt ŵt Combiig these observatios with 3.4 we get 3.5 Next, observe that 3.6 e βt ρ 1 e βt w 0. ŵ tx 1 tẋ t + e βt ρ 1 e βt w 0 RX ta T. A ẋ t = A N ẋ N t. Usig the otatio itroduced before the statemet of the lemma, 3.5 ad 3.6 ca be rewritte as follows: x t + e βt ρ 1 e βt u t RÃT, à t x t = A N ẋ N t. Let Ã+ t deote the pseudo-iverse of à t [3]. For example, if rakãt=m, the Ã+ t = ÃT t à tãt t 1. The, PR à T := Ã+ tãt is the orthogoal projectio matrix oto RÃT ad P N à := I Ã+ tãt is the orthogoal projectio matrix oto N à [3]. From 3.8 we obtai P R à T x t = Ã+ tãt x t = Ã+ ta N ẋ N t, ad from 3.7, usig the fact that RÃT ad N à are orthogoal to each other, we get P N à x t = e βt ρ 1 e βt I Ã+ tãt u t. Combiig, we have x t = P R à T x t + P N à x t = Ã+ ta N ẋ N t e βt ρ 1 e βt I Ã+ tãt u t, which gives 3.1. To determie the covergece directios of the trajectories, we eed to study the relative covergece speeds of ẋ t, ẋ N t, etc. Thus, we compute limits of some of the expressios that appear i equatios 3.1 ad 3.2: lim ŵt = et w 0 t e, lim ŵ N t = et w 0 t e N, lim D et w 0 t = t X 1 et w 0, lim D N t = t S N 1, lim à t = t e T A X w, lim GN t = 0 t e T G N SN, w 0 lim u t = t e T w 0, lim u N t = w 0 t e T w N 0. w 0
8 8 REHA H. TÜTÜNCÜ Lemma lim à + t t = et w 0 A X + lim G + t N t = et w 0 G N SN +. Proof. This result about the limitig properties of the pseudo-iverses is a immediate cosequece of Lemma 2.3 i [4] ad equatios Differetiatig the idetity we obtai 3.15 xtst = wt = e t w 0 + hte, xtṡt + ẋtst = e t w 0 + ḣte = e t w 0 et w 0 1 β e 1 βt e + et w 0 e t e. Next, we will establish that ẋ N t ad ṡ t coverge to zero o slower tha exp{ 1 βt}. For this purpose, we cosider the ormalized directio vectors ˆx, ŝ which are defied as follows: 3.16 ˆxt = exp {1 β t} ẋt, ad ŝt = exp {1 β t} ṡt. From 3.15 it follows that 3.17 xtŝt + ˆxtst = et w 0 1 β e + e βt e T w 0 e w 0. The expressio o the right-had-side of 3.17 is clearly bouded. With some more work, we have the followig coclusio: Lemma 3.3. Let ˆxt, ŝt be as i 3.16 ad assume that ρ >. The, ˆx N t, ŝ t remais bouded as t teds to. Proof. We will prove that xtŝt ad ˆxtst remai bouded as t. The, sice x t ad s N t coverge to x > 0 ad s N > 0 respectively, ad therefore, remai bouded away from zero, we ca coclude that ˆx N t ad ŝ t remai bouded. Sice the right-had-side of 3.17 is bouded as t teds to, it is sufficiet to show that [xtŝt] T [ˆxtst] remais bouded below to coclude that both xtŝt ad ˆxtst have bouded orms as t. Let vt = x 1 2 ts 1 2 t = w 1 2 t ad δt = x 1 2 ts 1 2 t. The, xtŝt = vtδtŝt ad ˆxtst = vtδ 1 tˆxt. Note that, [δtŝt] T [ δ 1 tˆxt ] = [δtṡt] T [ δ 1 tẋt ] = 0. Let V t = diagvt, t = diagδt, ad W 0 = diagw 0. The, V 2 t = XtSt = et w 0 e 1 βt e t I+ e t W 0. Now, [xtŝt] T [ˆxtst] = [vtδtŝt] T [ vtδ 1 tˆxt ] = [δtŝt] T V 2 t [ δ 1 tˆxt ]
9 3.18 POTENTIAL REDUCTION TRAJECTORIES 9 = e 21 βt [δtṡt] T V 2 t [ δ 1 tẋt ] = et w 0 e 1 βt e 1 2βt [δtṡt] T [ δ 1 tẋt ] +e 1 2βt [δtṡt] T [ W 0 δ 1 tẋt ] [ ] T [ ] = e βt e 1 βt/2 δtṡt W0 e 1 βt/2 δ 1 tẋt. Recall from 3.9 that lim t ŵ j t = lim t e 1 βt w j t = et w 0, j. Therefore, we have lim t ŵj t = lim t e 1 βt/2 v j t = e T w 0 ad defiig ṽt = e 1 βt/2 vt vtt vt v 1 t, ρ e we have lim t ṽ j t = 1 β T w 0. Now, recallig equatio 2.8 we observe that the vectors e 1 βt/2 δ 1 tẋt ad e 1 βt/2 δtṡt are orthogoal projectios of the vector ṽt ito the ull space of A t ad rage space of [A t] T, respectively. Sice we showed that the vector ṽt is coverget as t teds to, both of these projectios coverge, ad therefore, the expressio i 3.18 coverges to zero. Thus, xtŝt ad ˆxtst have bouded orms as t. It is iterestig that the coclusio of the lemma above holds for ay ρ 0. A alterative proof of Lemma 3.3 ca be obtaied usig the proof techique i [5]. Combiig equatio 3.1, Lemmas 3.2 ad 3.3 we obtai the followig result: Lemma 3.4. Let ˆxt, ŝt be as i 3.16 ad assume that < ρ 2. The, ˆxt, ŝt remais bouded as t teds to. Proof. From 3.1 we have that 3.19 D tˆx t = Ã+ ta N ˆx N t e1 2βt ρ 1 e βt I Ã+ tãt u t. Whe, ρ 2, the factor e1 2βt is coverget as t teds to. Now, usig Lemma ρ1 e βt 3.2 ad the equatios , we coclude that the secod term i the righthad-side of the equatio above remais bouded. Combiig this observatio with the the fact that ˆx N t remais bouded as t teds to, we obtai that D tˆx t remais bouded. Usig 3.10 we coclude that ˆx t is also bouded as t teds to. The fact that ŝt is bouded follows similarly. Now, the followig two results are easy to prove: Theorem 3.5. Let ˆxt, ŝt be as i The, we have that lim t ˆx N t ad lim t ŝ t exist ad satisfy the followig equatios: lim ˆx N t = et w 0 t 1 β s N 1, lim ŝt = et w 0 t 1 β x 1. Proof. From 3.17 we have that x tŝ t + ˆx ts t = et w 0 e 1 β e + e βt T w 0 e w 0.
10 10 REHA H. TÜTÜNCÜ Takig the limit o the right-had-side as t we obtai et w 0 1 βe. Sice s t 0 ad ˆx t is bouded, we must the have that x tŝ t coverges to et w 0 1 βe. Sice x t x, it follows that lim t ŝ t exists ad satisfies The correspodig result for ˆx N t follows idetically. Let π = X ξ = X A X + A N s N 1, I A X + A X w 0, σ N = SN G N SN + G x 1 π N = SN I G N SN + G N SN w N 0. Observe that π = 0 if ad oly if w 0 RX AT which holds, for example, whe x 0, s 0 is o the cetral path ad w0 = µe for some µ > 0 the observatio that e RX AT follows easily from the optimality of x for the first problem i 2.5. Similarly, π N = 0 if ad oly if w N 0 RSN GT N. Theorem 3.6. Let ˆxt, ŝt be as i 3.16 ad assume that ρ 2. The, we have that lim t ˆx t ad lim t ŝ N t exist. Whe ρ < 2 we have the followig idetities: lim ˆx t = et w 0 t 1 β ξ, lim t ŝn t = et w 0 1 β σ N. Whe ρ = 2, the followig equatios hold: lim ˆx t = et w 0 t 2 ξ 2e T w 0 π, lim t ŝn t = et w 0 2 σ N 2e T w 0 π N. Proof. Recall equatio Whe ρ < 2, the secod term o the right-hadside coverges to zero sice e 1 2βt teds to zero ad everythig else is bouded. Thus, usig 3.10 ad 3.11 we have lim t ˆx t = X A X + A N lim t ˆx N t ad 3.22 is obtaied usig Theorem 3.5. Similarly, oe obtais Whe ρ = 2, the factor i frot of the secod term i 3.19 coverges to the positive costat β = 1 2. Therefore, usig Theorem 3.5 ad equatios we get 3.24 ad Limits of the ormalized vectors ẋt ẋt, ṡt ṡt are obtaied immediately from Theorems 3.5 ad 3.6: Corollary 3.7. Let xt, st for t 0 deote the solutio of the ODE 2.9 with the iitial coditio x0, s0 = x 0, s 0 with x 0, s 0 F 0 ad assume that ρ 2. All trajectories of this form satisfy the followig equatios: 3.26 where 3.27 q P = [ [ q P = ξ lim t ξ s N 1 e T w 0 2 π s N 1 ẋt ẋt = q P q P, lim t ṡt ṡt = ] [ x, ad q D = 1 ], ad q D = [ σ N x 1 σ N q D q D ], if ρ < 2, e T w 0 2 πn ], if ρ = 2.
11 POTENTIAL REDUCTION TRAJECTORIES 11 Whe ρ = 2 the TTY potetial-fuctio Φ ρ x, y is a homogeeous fuctio ad exp{φ ρ x, y} is a covex fuctio for all ρ 2 [17]. The value 2 also represets a threshold value for the covergece behavior of the KMY trajectories. Whe ρ = 2 the directio of covergece depeds o the iitial poit x 0, s 0 F 0 as idicated by the appearace of the w 0 = x 0 s 0 terms i the formulas. We ote that, whe ρ < 2 the asymptotic directio of covergece does ot deped o the iitial poit ad is idetical to that of the cetral path. Therefore, whe ρ < 2 all trajectories of the vector field give by the search directio of the Kojima-Mizuo-Yoshise s primaldual potetial-reductio algorithm coverge to the aalytic ceter of the optimal face tagetially to the cetral path. We show below that the asymptotic behavior of the trajectories is sigificatly differet whe ρ > 2: Theorem 3.8. Let xt, st for t 0 deote the solutio of the ODE 2.9 with the iitial coditio x0, s0 = x 0, s 0 ad assume that ρ > 2. Defie 3.28 xt = e βt ẋt, ad st = e βt ṡt. If π 0 ad π N 0, the we have that lim t xt ad lim t st exist ad satisfy the followig equatios: 3.29 where lim t xt xt = q P = q P, ad lim q P t st st = [ ] [ ] π 0, ad q 0 D =. N π N q D q D, Proof. From 3.1 we have that D t x t = Ã+ ta 3.30 N x N t ρ 1 e βt I Ã+ tãt u t. Note that, x N t = e 1 2βtˆx N t. Sice ˆx N t is bouded ad 1 2β < 0, we coclude that x N t 0. Therefore, usig equatios 3.30 ad we observe that x t coverges to a positive multiple of π 0 ad immediately obtai the first equatio i The secod idetity i 3.29 is obtaied similarly. This fial theorem idicates that whe ρ > 2, most trajectories associated with the Kojima-Mizuo-Yoshise algorithm coverge to the aalytic ceter of the optimal face tagetially to the optimal face ad their directio of covergece depeds o the iitial poit. I the exceptioal case of π = 0 or π N = 0 for example, whe x 0, s 0 is o the cetral path, the last term i 3.30 o loger domiates the righthad-side ad i such cases we cojecture that the trajectory coverges tagetially to the cetral path. We coclude by otig the similarity of our asymptotic results to those of Moteiro [12]. I his aalysis of the trajectories based o primal-oly potetial-reductio algorithms, Moteiro also fids that there is a threshold value of the potetial fuctio parameter that leads to differet asymptotic behavior. I his case, this threshold value is 2 N rather tha 2, where N deotes the cardiality of the set N from the optimal partitio of {1,..., }. Just like our case, whe the potetial fuctio
12 12 REHA H. TÜTÜNCÜ parameter is above this value, his trajectories coverge tagetially to the cetral path ad below the threshold the covergece is tagetial to the optimal face. ACKNOWLEDGMENT The coclusio of Lemma 3.3 was stated without a proof i a earlier draft of this paper. The author would like to thak Margaréta Halická whose careful readig led to the correctio of this omissio. She also provided a alterative argumet for the proof. The author would also like to thak a aoymous referee for brigig refereces [7, 9, 10] to his attetio ad makig several costructive commets. Results i these refereces helped make this a more cocise paper. REFERENCES [1] I. Adler ad R. D. C. Moteiro, Limitig behavior of the affie scalig cotiuous trajectories for liear programmig problems, Mathematical Programmig, pp [2] K. M. Astreicher, Potetial reductio algorithms, i: T. Terlaky, ed., Iterior poit methods of mathematical programmig, Kluwer Academic Publishers, Dordrecht, Netherlads, 1996 pp [3] G. H. Golub ad C. F. Va Loa, Matrix Computatios, 2d ed., The Johs Hopkis Uiversity Press, altimore, [4] O. Güler, Limitig behavior of weighted cetral paths i liear programmig, Mathematical Programmig, pp [5] M. Halická, Two simple proofs for aalyticity of the cetral path, Operatios Research Letters, pp [6] M. W. Hirsch ad S. Smale, Differetial equatios, dyamical systems, ad liear algebra Academic Press, New York, [7] M. Kojima, N. Megiddo, T. Noma, ad A. Yoshise, A uified approach to iterior poit algorithms for liear complemetarity problems, Vol. 538 of Lecture Notes i Computer Sciece, Spriger Verlag, erli, Germay, [8] M. Kojima, S. Mizuo, ad A. Yoshise, A O L iteratio potetial reductio algorithm for liear complemetarity problems, Mathematical Programmig, pp [9] L. McLide, A aalogue of Moreau s proximatio theorem, with applicatios to the oliear complemetarity problem, Pacific Joural of Mathematics, pp [10] L. McLide, The complemetarity problem for maximal mootoe multifuctios, i: R. W. Cottle, F. Giaessi, ad J.-L. Lios, eds., Variatioal Iequalities ad Complemetarity Problems, Wiley, New York, 1980 pp [11] R. D. C. Moteiro, Covergece ad boudary behavior of the projective scalig trajectories for liear programmig, Mathematics of Operatios Research, pp [12] R. D. C. Moteiro, O the cotiuous trajectories for potetial reductio algorithms for liear programmig, Mathematics of Operatios Research, pp [13] Yu. Nesterov, Log-step strategies i iterior-poit primal-dual methods, Mathematical Programmig, pp [14] K. Taabe, Cetered Newto method for mathematical programmig, i: M. Iri ad K. Yajima, eds., Lecture Notes i Cotrol ad Iformatio Scieces, Spriger-Verlag, erli, 1988 pp [15] M. J. Todd, Potetial-reductio methods i mathematical programmig, Mathematical Programmig, [16] M. J. Todd ad Y. Ye, A cetered projective algorithm for liear programmig, Mathematics of Operatios Research, [17] R. H. Tütücü, A primal-dual variat of the Iri-Imai algorithm for liear programmig, Mathematics of Operatios Research, , [18] S. Wright, Primal-dual iterior-poit methods SIAM, Philadelphia, 1997.
Continuous Trajectories for Primal-Dual Potential-Reduction Methods
Cotiuous Trajectories for Primal-Dual Potetial-Reductio Methods Reha H. Tütücü September 7, 2001 Abstract This article cosiders cotiuous trajectories of the vector fields iduced by two differet primal-dual
More informationOptimization Methods MIT 2.098/6.255/ Final exam
Optimizatio Methods MIT 2.098/6.255/15.093 Fial exam Date Give: December 19th, 2006 P1. [30 pts] Classify the followig statemets as true or false. All aswers must be well-justified, either through a short
More information2 The LCP problem and preliminary discussion
Iteratioal Mathematical Forum, Vol. 6, 011, o. 6, 191-196 Polyomial Covergece of a Predictor-Corrector Iterior-Poit Algorithm for LCP Feixiag Che College of Mathematics ad Computer Sciece, Chogqig Three
More informationA Hadamard-type lower bound for symmetric diagonally dominant positive matrices
A Hadamard-type lower boud for symmetric diagoally domiat positive matrices Christopher J. Hillar, Adre Wibisoo Uiversity of Califoria, Berkeley Jauary 7, 205 Abstract We prove a ew lower-boud form of
More informationConvergence of random variables. (telegram style notes) P.J.C. Spreij
Covergece of radom variables (telegram style otes).j.c. Spreij this versio: September 6, 2005 Itroductio As we kow, radom variables are by defiitio measurable fuctios o some uderlyig measurable space
More informationIntroduction to Optimization Techniques. How to Solve Equations
Itroductio to Optimizatio Techiques How to Solve Equatios Iterative Methods of Optimizatio Iterative methods of optimizatio Solutio of the oliear equatios resultig form a optimizatio problem is usually
More informationlim za n n = z lim a n n.
Lecture 6 Sequeces ad Series Defiitio 1 By a sequece i a set A, we mea a mappig f : N A. It is customary to deote a sequece f by {s } where, s := f(). A sequece {z } of (complex) umbers is said to be coverget
More information1 Duality revisited. AM 221: Advanced Optimization Spring 2016
AM 22: Advaced Optimizatio Sprig 206 Prof. Yaro Siger Sectio 7 Wedesday, Mar. 9th Duality revisited I this sectio, we will give a slightly differet perspective o duality. optimizatio program: f(x) x R
More information15.081J/6.251J Introduction to Mathematical Programming. Lecture 21: Primal Barrier Interior Point Algorithm
508J/65J Itroductio to Mathematical Programmig Lecture : Primal Barrier Iterior Poit Algorithm Outlie Barrier Methods Slide The Cetral Path 3 Approximatig the Cetral Path 4 The Primal Barrier Algorithm
More informationSupplemental Material: Proofs
Proof to Theorem Supplemetal Material: Proofs Proof. Let be the miimal umber of traiig items to esure a uique solutio θ. First cosider the case. It happes if ad oly if θ ad Rak(A) d, which is a special
More informationA Primal-Dual Interior-Point Filter Method for Nonlinear Semidefinite Programming
The 7th Iteratioal Symposium o Operatios Research ad Its Applicatios (ISORA 8) Lijiag, Chia, October 31 Novemver 3, 28 Copyright 28 ORSC & APORC, pp. 112 118 A Primal-Dual Iterior-Poit Filter Method for
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 Fuctioal Law of Large Numbers. Costructio of the Wieer Measure Cotet. 1. Additioal techical results o weak covergece
More informationChapter 7 Isoperimetric problem
Chapter 7 Isoperimetric problem Recall that the isoperimetric problem (see the itroductio its coectio with ido s proble) is oe of the most classical problem of a shape optimizatio. It ca be formulated
More informationw (1) ˆx w (1) x (1) /ρ and w (2) ˆx w (2) x (2) /ρ.
2 5. Weighted umber of late jobs 5.1. Release dates ad due dates: maximimizig the weight of o-time jobs Oce we add release dates, miimizig the umber of late jobs becomes a sigificatly harder problem. For
More informationDefinition 4.2. (a) A sequence {x n } in a Banach space X is a basis for X if. unique scalars a n (x) such that x = n. a n (x) x n. (4.
4. BASES I BAACH SPACES 39 4. BASES I BAACH SPACES Sice a Baach space X is a vector space, it must possess a Hamel, or vector space, basis, i.e., a subset {x γ } γ Γ whose fiite liear spa is all of X ad
More informationFeedback in Iterative Algorithms
Feedback i Iterative Algorithms Charles Byre (Charles Byre@uml.edu), Departmet of Mathematical Scieces, Uiversity of Massachusetts Lowell, Lowell, MA 01854 October 17, 2005 Abstract Whe the oegative system
More informationInfinite Sequences and Series
Chapter 6 Ifiite Sequeces ad Series 6.1 Ifiite Sequeces 6.1.1 Elemetary Cocepts Simply speakig, a sequece is a ordered list of umbers writte: {a 1, a 2, a 3,...a, a +1,...} where the elemets a i represet
More informationComplex Analysis Spring 2001 Homework I Solution
Complex Aalysis Sprig 2001 Homework I Solutio 1. Coway, Chapter 1, sectio 3, problem 3. Describe the set of poits satisfyig the equatio z a z + a = 2c, where c > 0 ad a R. To begi, we see from the triagle
More informationThe log-behavior of n p(n) and n p(n)/n
Ramauja J. 44 017, 81-99 The log-behavior of p ad p/ William Y.C. Che 1 ad Ke Y. Zheg 1 Ceter for Applied Mathematics Tiaji Uiversity Tiaji 0007, P. R. Chia Ceter for Combiatorics, LPMC Nakai Uivercity
More informationLecture 7: October 18, 2017
Iformatio ad Codig Theory Autum 207 Lecturer: Madhur Tulsiai Lecture 7: October 8, 207 Biary hypothesis testig I this lecture, we apply the tools developed i the past few lectures to uderstad the problem
More informationCHAPTER I: Vector Spaces
CHAPTER I: Vector Spaces Sectio 1: Itroductio ad Examples This first chapter is largely a review of topics you probably saw i your liear algebra course. So why cover it? (1) Not everyoe remembers everythig
More informationLecture Notes for Analysis Class
Lecture Notes for Aalysis Class Topological Spaces A topology for a set X is a collectio T of subsets of X such that: (a) X ad the empty set are i T (b) Uios of elemets of T are i T (c) Fiite itersectios
More information62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 +
62. Power series Defiitio 16. (Power series) Give a sequece {c }, the series c x = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + is called a power series i the variable x. The umbers c are called the coefficiets of
More informationProblem Cosider the curve give parametrically as x = si t ad y = + cos t for» t» ß: (a) Describe the path this traverses: Where does it start (whe t =
Mathematics Summer Wilso Fial Exam August 8, ANSWERS Problem 1 (a) Fid the solutio to y +x y = e x x that satisfies y() = 5 : This is already i the form we used for a first order liear differetial equatio,
More informationChapter 3. Strong convergence. 3.1 Definition of almost sure convergence
Chapter 3 Strog covergece As poited out i the Chapter 2, there are multiple ways to defie the otio of covergece of a sequece of radom variables. That chapter defied covergece i probability, covergece i
More informationCHAPTER 10 INFINITE SEQUENCES AND SERIES
CHAPTER 10 INFINITE SEQUENCES AND SERIES 10.1 Sequeces 10.2 Ifiite Series 10.3 The Itegral Tests 10.4 Compariso Tests 10.5 The Ratio ad Root Tests 10.6 Alteratig Series: Absolute ad Coditioal Covergece
More informationChapter 6 Infinite Series
Chapter 6 Ifiite Series I the previous chapter we cosidered itegrals which were improper i the sese that the iterval of itegratio was ubouded. I this chapter we are goig to discuss a topic which is somewhat
More informationComputation of Error Bounds for P-matrix Linear Complementarity Problems
Mathematical Programmig mauscript No. (will be iserted by the editor) Xiaoju Che Shuhuag Xiag Computatio of Error Bouds for P-matrix Liear Complemetarity Problems Received: date / Accepted: date Abstract
More informationTHE ASYMPTOTIC COMPLEXITY OF MATRIX REDUCTION OVER FINITE FIELDS
THE ASYMPTOTIC COMPLEXITY OF MATRIX REDUCTION OVER FINITE FIELDS DEMETRES CHRISTOFIDES Abstract. Cosider a ivertible matrix over some field. The Gauss-Jorda elimiatio reduces this matrix to the idetity
More informationSequences and Series of Functions
Chapter 6 Sequeces ad Series of Fuctios 6.1. Covergece of a Sequece of Fuctios Poitwise Covergece. Defiitio 6.1. Let, for each N, fuctio f : A R be defied. If, for each x A, the sequece (f (x)) coverges
More informationSome remarks for codes and lattices over imaginary quadratic
Some remarks for codes ad lattices over imagiary quadratic fields Toy Shaska Oaklad Uiversity, Rochester, MI, USA. Caleb Shor Wester New Eglad Uiversity, Sprigfield, MA, USA. shaska@oaklad.edu Abstract
More information6.3 Testing Series With Positive Terms
6.3. TESTING SERIES WITH POSITIVE TERMS 307 6.3 Testig Series With Positive Terms 6.3. Review of what is kow up to ow I theory, testig a series a i for covergece amouts to fidig the i= sequece of partial
More informationMath 451: Euclidean and Non-Euclidean Geometry MWF 3pm, Gasson 204 Homework 3 Solutions
Math 451: Euclidea ad No-Euclidea Geometry MWF 3pm, Gasso 204 Homework 3 Solutios Exercises from 1.4 ad 1.5 of the otes: 4.3, 4.10, 4.12, 4.14, 4.15, 5.3, 5.4, 5.5 Exercise 4.3. Explai why Hp, q) = {x
More informationRiesz-Fischer Sequences and Lower Frame Bounds
Zeitschrift für Aalysis ud ihre Aweduge Joural for Aalysis ad its Applicatios Volume 1 (00), No., 305 314 Riesz-Fischer Sequeces ad Lower Frame Bouds P. Casazza, O. Christese, S. Li ad A. Lider Abstract.
More informationMachine Learning for Data Science (CS 4786)
Machie Learig for Data Sciece CS 4786) Lecture & 3: Pricipal Compoet Aalysis The text i black outlies high level ideas. The text i blue provides simple mathematical details to derive or get to the algorithm
More informationNumerical Method for Blasius Equation on an infinite Interval
Numerical Method for Blasius Equatio o a ifiite Iterval Alexader I. Zadori Omsk departmet of Sobolev Mathematics Istitute of Siberia Brach of Russia Academy of Scieces, Russia zadori@iitam.omsk.et.ru 1
More informationOn forward improvement iteration for stopping problems
O forward improvemet iteratio for stoppig problems Mathematical Istitute, Uiversity of Kiel, Ludewig-Mey-Str. 4, D-24098 Kiel, Germay irle@math.ui-iel.de Albrecht Irle Abstract. We cosider the optimal
More informationOptimally Sparse SVMs
A. Proof of Lemma 3. We here prove a lower boud o the umber of support vectors to achieve geeralizatio bouds of the form which we cosider. Importatly, this result holds ot oly for liear classifiers, but
More informationLinear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d
Liear regressio Daiel Hsu (COMS 477) Maximum likelihood estimatio Oe of the simplest liear regressio models is the followig: (X, Y ),..., (X, Y ), (X, Y ) are iid radom pairs takig values i R d R, ad Y
More informationA 2nTH ORDER LINEAR DIFFERENCE EQUATION
A 2TH ORDER LINEAR DIFFERENCE EQUATION Doug Aderso Departmet of Mathematics ad Computer Sciece, Cocordia College Moorhead, MN 56562, USA ABSTRACT: We give a formulatio of geeralized zeros ad (, )-discojugacy
More informationSequences. Notation. Convergence of a Sequence
Sequeces A sequece is essetially just a list. Defiitio (Sequece of Real Numbers). A sequece of real umbers is a fuctio Z (, ) R for some real umber. Do t let the descriptio of the domai cofuse you; it
More informationLecture 19: Convergence
Lecture 19: Covergece Asymptotic approach I statistical aalysis or iferece, a key to the success of fidig a good procedure is beig able to fid some momets ad/or distributios of various statistics. I may
More informationJournal of Multivariate Analysis. Superefficient estimation of the marginals by exploiting knowledge on the copula
Joural of Multivariate Aalysis 102 (2011) 1315 1319 Cotets lists available at ScieceDirect Joural of Multivariate Aalysis joural homepage: www.elsevier.com/locate/jmva Superefficiet estimatio of the margials
More informationPAijpam.eu ON TENSOR PRODUCT DECOMPOSITION
Iteratioal Joural of Pure ad Applied Mathematics Volume 103 No 3 2015, 537-545 ISSN: 1311-8080 (prited versio); ISSN: 1314-3395 (o-lie versio) url: http://wwwijpameu doi: http://dxdoiorg/1012732/ijpamv103i314
More informationAdvanced Analysis. Min Yan Department of Mathematics Hong Kong University of Science and Technology
Advaced Aalysis Mi Ya Departmet of Mathematics Hog Kog Uiversity of Sciece ad Techology September 3, 009 Cotets Limit ad Cotiuity 7 Limit of Sequece 8 Defiitio 8 Property 3 3 Ifiity ad Ifiitesimal 8 4
More informationTHE SOLUTION OF NONLINEAR EQUATIONS f( x ) = 0.
THE SOLUTION OF NONLINEAR EQUATIONS f( ) = 0. Noliear Equatio Solvers Bracketig. Graphical. Aalytical Ope Methods Bisectio False Positio (Regula-Falsi) Fied poit iteratio Newto Raphso Secat The root of
More informationFLOOR AND ROOF FUNCTION ANALOGS OF THE BELL NUMBERS. H. W. Gould Department of Mathematics, West Virginia University, Morgantown, WV 26506, USA
INTEGERS: ELECTRONIC JOURNAL OF COMBINATORIAL NUMBER THEORY 7 (2007), #A58 FLOOR AND ROOF FUNCTION ANALOGS OF THE BELL NUMBERS H. W. Gould Departmet of Mathematics, West Virgiia Uiversity, Morgatow, WV
More informationEntropy Rates and Asymptotic Equipartition
Chapter 29 Etropy Rates ad Asymptotic Equipartitio Sectio 29. itroduces the etropy rate the asymptotic etropy per time-step of a stochastic process ad shows that it is well-defied; ad similarly for iformatio,
More informationSolutions to home assignments (sketches)
Matematiska Istitutioe Peter Kumli 26th May 2004 TMA401 Fuctioal Aalysis MAN670 Applied Fuctioal Aalysis 4th quarter 2003/2004 All documet cocerig the course ca be foud o the course home page: http://www.math.chalmers.se/math/grudutb/cth/tma401/
More informationProbability 2 - Notes 10. Lemma. If X is a random variable and g(x) 0 for all x in the support of f X, then P(g(X) 1) E[g(X)].
Probability 2 - Notes 0 Some Useful Iequalities. Lemma. If X is a radom variable ad g(x 0 for all x i the support of f X, the P(g(X E[g(X]. Proof. (cotiuous case P(g(X Corollaries x:g(x f X (xdx x:g(x
More informationVector Quantization: a Limiting Case of EM
. Itroductio & defiitios Assume that you are give a data set X = { x j }, j { 2,,, }, of d -dimesioal vectors. The vector quatizatio (VQ) problem requires that we fid a set of prototype vectors Z = { z
More informationNotes on iteration and Newton s method. Iteration
Notes o iteratio ad Newto s method Iteratio Iteratio meas doig somethig over ad over. I our cotet, a iteratio is a sequece of umbers, vectors, fuctios, etc. geerated by a iteratio rule of the type 1 f
More informationMath 312 Lecture Notes One Dimensional Maps
Math 312 Lecture Notes Oe Dimesioal Maps Warre Weckesser Departmet of Mathematics Colgate Uiversity 21-23 February 25 A Example We begi with the simplest model of populatio growth. Suppose, for example,
More informationMathematical Methods for Physics and Engineering
Mathematical Methods for Physics ad Egieerig Lecture otes Sergei V. Shabaov Departmet of Mathematics, Uiversity of Florida, Gaiesville, FL 326 USA CHAPTER The theory of covergece. Numerical sequeces..
More informationSeunghee Ye Ma 8: Week 5 Oct 28
Week 5 Summary I Sectio, we go over the Mea Value Theorem ad its applicatios. I Sectio 2, we will recap what we have covered so far this term. Topics Page Mea Value Theorem. Applicatios of the Mea Value
More informationProperties of Fuzzy Length on Fuzzy Set
Ope Access Library Joural 206, Volume 3, e3068 ISSN Olie: 2333-972 ISSN Prit: 2333-9705 Properties of Fuzzy Legth o Fuzzy Set Jehad R Kider, Jaafar Imra Mousa Departmet of Mathematics ad Computer Applicatios,
More informationWe are mainly going to be concerned with power series in x, such as. (x)} converges - that is, lims N n
Review of Power Series, Power Series Solutios A power series i x - a is a ifiite series of the form c (x a) =c +c (x a)+(x a) +... We also call this a power series cetered at a. Ex. (x+) is cetered at
More informationNEW FAST CONVERGENT SEQUENCES OF EULER-MASCHERONI TYPE
UPB Sci Bull, Series A, Vol 79, Iss, 207 ISSN 22-7027 NEW FAST CONVERGENT SEQUENCES OF EULER-MASCHERONI TYPE Gabriel Bercu We itroduce two ew sequeces of Euler-Mascheroi type which have fast covergece
More informationLecture XVI - Lifting of paths and homotopies
Lecture XVI - Liftig of paths ad homotopies I the last lecture we discussed the liftig problem ad proved that the lift if it exists is uiquely determied by its value at oe poit. I this lecture we shall
More informationHomework Set #3 - Solutions
EE 15 - Applicatios of Covex Optimizatio i Sigal Processig ad Commuicatios Dr. Adre Tkaceko JPL Third Term 11-1 Homework Set #3 - Solutios 1. a) Note that x is closer to x tha to x l i the Euclidea orm
More informationResearch Article An Improved Predictor-Corrector Interior-Point Algorithm for Linear Complementarity Problems with O nl -Iteration Complexity
Joural of Applied Mathematics Volume 0, Article ID 3409, pages doi:0.55/0/3409 Research Article A Improved Predictor-Corrector Iterior-Poit Algorithm for Liear Complemetarity Problems with O L-Iteratio
More informationA constructive analysis of convex-valued demand correspondence for weakly uniformly rotund and monotonic preference
MPRA Muich Persoal RePEc Archive A costructive aalysis of covex-valued demad correspodece for weakly uiformly rotud ad mootoic preferece Yasuhito Taaka ad Atsuhiro Satoh. May 04 Olie at http://mpra.ub.ui-mueche.de/55889/
More informationTechnical Proofs for Homogeneity Pursuit
Techical Proofs for Homogeeity Pursuit bstract This is the supplemetal material for the article Homogeeity Pursuit, submitted for publicatio i Joural of the merica Statistical ssociatio. B Proofs B. Proof
More informationOn Nonsingularity of Saddle Point Matrices. with Vectors of Ones
Iteratioal Joural of Algebra, Vol. 2, 2008, o. 4, 197-204 O Nosigularity of Saddle Poit Matrices with Vectors of Oes Tadeusz Ostrowski Istitute of Maagemet The State Vocatioal Uiversity -400 Gorzów, Polad
More informationReview Problems 1. ICME and MS&E Refresher Course September 19, 2011 B = C = AB = A = A 2 = A 3... C 2 = C 3 = =
Review Problems ICME ad MS&E Refresher Course September 9, 0 Warm-up problems. For the followig matrices A = 0 B = C = AB = 0 fid all powers A,A 3,(which is A times A),... ad B,B 3,... ad C,C 3,... Solutio:
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 6 9/23/2013. Brownian motion. Introduction
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/5.070J Fall 203 Lecture 6 9/23/203 Browia motio. Itroductio Cotet.. A heuristic costructio of a Browia motio from a radom walk. 2. Defiitio ad basic properties
More informationb i u x i U a i j u x i u x j
M ath 5 2 7 Fall 2 0 0 9 L ecture 1 9 N ov. 1 6, 2 0 0 9 ) S ecod- Order Elliptic Equatios: Weak S olutios 1. Defiitios. I this ad the followig two lectures we will study the boudary value problem Here
More information7.1 Convergence of sequences of random variables
Chapter 7 Limit Theorems Throughout this sectio we will assume a probability space (, F, P), i which is defied a ifiite sequece of radom variables (X ) ad a radom variable X. The fact that for every ifiite
More informationA RANK STATISTIC FOR NON-PARAMETRIC K-SAMPLE AND CHANGE POINT PROBLEMS
J. Japa Statist. Soc. Vol. 41 No. 1 2011 67 73 A RANK STATISTIC FOR NON-PARAMETRIC K-SAMPLE AND CHANGE POINT PROBLEMS Yoichi Nishiyama* We cosider k-sample ad chage poit problems for idepedet data i a
More information(A sequence also can be thought of as the list of function values attained for a function f :ℵ X, where f (n) = x n for n 1.) x 1 x N +k x N +4 x 3
MATH 337 Sequeces Dr. Neal, WKU Let X be a metric space with distace fuctio d. We shall defie the geeral cocept of sequece ad limit i a metric space, the apply the results i particular to some special
More informationMulti parameter proximal point algorithms
Multi parameter proximal poit algorithms Ogaeditse A. Boikayo a,b,, Gheorghe Moroşau a a Departmet of Mathematics ad its Applicatios Cetral Europea Uiversity Nador u. 9, H-1051 Budapest, Hugary b Departmet
More informationIterative Techniques for Solving Ax b -(3.8). Assume that the system has a unique solution. Let x be the solution. Then x A 1 b.
Iterative Techiques for Solvig Ax b -(8) Cosider solvig liear systems of them form: Ax b where A a ij, x x i, b b i Assume that the system has a uique solutio Let x be the solutio The x A b Jacobi ad Gauss-Seidel
More informationSimilarity Solutions to Unsteady Pseudoplastic. Flow Near a Moving Wall
Iteratioal Mathematical Forum, Vol. 9, 04, o. 3, 465-475 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/0.988/imf.04.48 Similarity Solutios to Usteady Pseudoplastic Flow Near a Movig Wall W. Robi Egieerig
More informationMATH301 Real Analysis (2008 Fall) Tutorial Note #7. k=1 f k (x) converges pointwise to S(x) on E if and
MATH01 Real Aalysis (2008 Fall) Tutorial Note #7 Sequece ad Series of fuctio 1: Poitwise Covergece ad Uiform Covergece Part I: Poitwise Covergece Defiitio of poitwise covergece: A sequece of fuctios f
More information7.1 Convergence of sequences of random variables
Chapter 7 Limit theorems Throughout this sectio we will assume a probability space (Ω, F, P), i which is defied a ifiite sequece of radom variables (X ) ad a radom variable X. The fact that for every ifiite
More informationAssignment 5: Solutions
McGill Uiversity Departmet of Mathematics ad Statistics MATH 54 Aalysis, Fall 05 Assigmet 5: Solutios. Let y be a ubouded sequece of positive umbers satisfyig y + > y for all N. Let x be aother sequece
More informationVariable selection in principal components analysis of qualitative data using the accelerated ALS algorithm
Variable selectio i pricipal compoets aalysis of qualitative data usig the accelerated ALS algorithm Masahiro Kuroda Yuichi Mori Masaya Iizuka Michio Sakakihara (Okayama Uiversity of Sciece) (Okayama Uiversity
More informationLECTURE 8: ORTHOGONALITY (CHAPTER 5 IN THE BOOK)
LECTURE 8: ORTHOGONALITY (CHAPTER 5 IN THE BOOK) Everythig marked by is ot required by the course syllabus I this lecture, all vector spaces is over the real umber R. All vectors i R is viewed as a colum
More information7 Sequences of real numbers
40 7 Sequeces of real umbers 7. Defiitios ad examples Defiitio 7... A sequece of real umbers is a real fuctio whose domai is the set N of atural umbers. Let s : N R be a sequece. The the values of s are
More informationCHAPTER 5. Theory and Solution Using Matrix Techniques
A SERIES OF CLASS NOTES FOR 2005-2006 TO INTRODUCE LINEAR AND NONLINEAR PROBLEMS TO ENGINEERS, SCIENTISTS, AND APPLIED MATHEMATICIANS DE CLASS NOTES 3 A COLLECTION OF HANDOUTS ON SYSTEMS OF ORDINARY DIFFERENTIAL
More informationBeurling Integers: Part 2
Beurlig Itegers: Part 2 Isomorphisms Devi Platt July 11, 2015 1 Prime Factorizatio Sequeces I the last article we itroduced the Beurlig geeralized itegers, which ca be represeted as a sequece of real umbers
More information3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense,
3. Z Trasform Referece: Etire Chapter 3 of text. Recall that the Fourier trasform (FT) of a DT sigal x [ ] is ω ( ) [ ] X e = j jω k = xe I order for the FT to exist i the fiite magitude sese, S = x [
More informationON WELLPOSEDNESS QUADRATIC FUNCTION MINIMIZATION PROBLEM ON INTERSECTION OF TWO ELLIPSOIDS * M. JA]IMOVI], I. KRNI] 1.
Yugoslav Joural of Operatios Research 1 (00), Number 1, 49-60 ON WELLPOSEDNESS QUADRATIC FUNCTION MINIMIZATION PROBLEM ON INTERSECTION OF TWO ELLIPSOIDS M. JA]IMOVI], I. KRNI] Departmet of Mathematics
More informationMA131 - Analysis 1. Workbook 3 Sequences II
MA3 - Aalysis Workbook 3 Sequeces II Autum 2004 Cotets 2.8 Coverget Sequeces........................ 2.9 Algebra of Limits......................... 2 2.0 Further Useful Results........................
More informationDefinitions and Theorems. where x are the decision variables. c, b, and a are constant coefficients.
Defiitios ad Theorems Remember the scalar form of the liear programmig problem, Miimize, Subject to, f(x) = c i x i a 1i x i = b 1 a mi x i = b m x i 0 i = 1,2,, where x are the decisio variables. c, b,
More informationDirichlet s Theorem on Arithmetic Progressions
Dirichlet s Theorem o Arithmetic Progressios Athoy Várilly Harvard Uiversity, Cambridge, MA 0238 Itroductio Dirichlet s theorem o arithmetic progressios is a gem of umber theory. A great part of its beauty
More informationComparison Study of Series Approximation. and Convergence between Chebyshev. and Legendre Series
Applied Mathematical Scieces, Vol. 7, 03, o. 6, 3-337 HIKARI Ltd, www.m-hikari.com http://d.doi.org/0.988/ams.03.3430 Compariso Study of Series Approimatio ad Covergece betwee Chebyshev ad Legedre Series
More informationProblem Set 2 Solutions
CS271 Radomess & Computatio, Sprig 2018 Problem Set 2 Solutios Poit totals are i the margi; the maximum total umber of poits was 52. 1. Probabilistic method for domiatig sets 6pts Pick a radom subset S
More informationProduct measures, Tonelli s and Fubini s theorems For use in MAT3400/4400, autumn 2014 Nadia S. Larsen. Version of 13 October 2014.
Product measures, Toelli s ad Fubii s theorems For use i MAT3400/4400, autum 2014 Nadia S. Larse Versio of 13 October 2014. 1. Costructio of the product measure The purpose of these otes is to preset the
More informationEfficient GMM LECTURE 12 GMM II
DECEMBER 1 010 LECTURE 1 II Efficiet The estimator depeds o the choice of the weight matrix A. The efficiet estimator is the oe that has the smallest asymptotic variace amog all estimators defied by differet
More information4. Partial Sums and the Central Limit Theorem
1 of 10 7/16/2009 6:05 AM Virtual Laboratories > 6. Radom Samples > 1 2 3 4 5 6 7 4. Partial Sums ad the Cetral Limit Theorem The cetral limit theorem ad the law of large umbers are the two fudametal theorems
More informationSummary and Discussion on Simultaneous Analysis of Lasso and Dantzig Selector
Summary ad Discussio o Simultaeous Aalysis of Lasso ad Datzig Selector STAT732, Sprig 28 Duzhe Wag May 4, 28 Abstract This is a discussio o the work i Bickel, Ritov ad Tsybakov (29). We begi with a short
More informationMath 61CM - Solutions to homework 3
Math 6CM - Solutios to homework 3 Cédric De Groote October 2 th, 208 Problem : Let F be a field, m 0 a fixed oegative iteger ad let V = {a 0 + a x + + a m x m a 0,, a m F} be the vector space cosistig
More informationAlgebra of Least Squares
October 19, 2018 Algebra of Least Squares Geometry of Least Squares Recall that out data is like a table [Y X] where Y collects observatios o the depedet variable Y ad X collects observatios o the k-dimesioal
More informationLimit distributions for products of sums
Statistics & Probability Letters 62 (23) 93 Limit distributios for products of sums Yogcheg Qi Departmet of Mathematics ad Statistics, Uiversity of Miesota-Duluth, Campus Ceter 4, 7 Uiversity Drive, Duluth,
More informationLinear Support Vector Machines
Liear Support Vector Machies David S. Roseberg The Support Vector Machie For a liear support vector machie (SVM), we use the hypothesis space of affie fuctios F = { f(x) = w T x + b w R d, b R } ad evaluate
More informationMath 155 (Lecture 3)
Math 55 (Lecture 3) September 8, I this lecture, we ll cosider the aswer to oe of the most basic coutig problems i combiatorics Questio How may ways are there to choose a -elemet subset of the set {,,,
More informationRandom assignment with integer costs
Radom assigmet with iteger costs Robert Parviaie Departmet of Mathematics, Uppsala Uiversity P.O. Box 480, SE-7506 Uppsala, Swede robert.parviaie@math.uu.se Jue 4, 200 Abstract The radom assigmet problem
More informationPAPER : IIT-JAM 2010
MATHEMATICS-MA (CODE A) Q.-Q.5: Oly oe optio is correct for each questio. Each questio carries (+6) marks for correct aswer ad ( ) marks for icorrect aswer.. Which of the followig coditios does NOT esure
More informationStatistical Inference Based on Extremum Estimators
T. Rotheberg Fall, 2007 Statistical Iferece Based o Extremum Estimators Itroductio Suppose 0, the true value of a p-dimesioal parameter, is kow to lie i some subset S R p : Ofte we choose to estimate 0
More information