Reverse Engineering TCP/IP-like Networks using Delay-Sensitive Utility Functions

Size: px
Start display at page:

Download "Reverse Engineering TCP/IP-like Networks using Delay-Sensitive Utility Functions"

Transcription

1 Reverse Engneerng TCP/IP-ke Networks usng Deay-Senstve Utty Functons John Pongsajapan and Steven H. Low Engneerng & Apped Scence, Catech Abstract TCP/IP can be nterpreted as a dstrbuted prmadua agorthm to maxmze aggregate utty over source rates. It has recenty been shown that an equbrum of TCP/IP, f t exsts, maxmzes the same deay-nsenstve utty over both source rates and routes, provded pure congeston prces are used as nk costs n the shortest-path cacuaton of IP. In practce, however, pure dynamc routng s never used and nk costs are weghted sums of both statc as we as dynamc components. In ths paper, we ntroduce deay-senstve utty functons and dentfy a cass of utty functons that such a TCP/IP equbrum optmzes. We exhbt some counter-ntutve propertes that any cass of deay-senstve utty functons optmzed by TCP/IP necessary possess. We prove a suffcent condton for goba stabty of routng updates for genera networks. We construct exampe networks that defy conventona wsdom on the effect of nk cost parameters on network stabty and utty. I. INTRODUCTION AND SUMMARY Any TCP congeston contro agorthm can be nterpreted as carryng out a dstrbuted prma-dua agorthm over the Internet to maxmze aggregate utty, see e.g. [12], [13], [17] [21], [25] for uncast and [4], [10], [25] for mutcast. A of these works assume that routng s gven and fxed at the tmescae of nterest, and TCP, together wth actve queue management AQM, attempt to maxmze aggregate utty over source rates. The paper [26] studes cross-ayer utty maxmzaton at the tmescae of route changes, many for the speca case of pure dynamc routng. In ths paper, we extend the resuts of [26] n severa ways. As n [26], we focus on the stuaton where a snge mnmum-cost route shortest path s seected for each sourcedestnaton par Secton II. Ths modes IP routng n the current Internet wthn an Autonomous Systems usng common routng protocos such as OSPF [22] 1 or RIP [8]. For jont congeston contro and routng optmzaton usng mutpe paths, see, e.g., [2], [5] [7], [9], [11], [12], [14], [15], [23]. Routng s typcay updated at a much sower tmescae than TCP AQM. We mode ths by assumng that TCP and AQM converge nstanty to equbrum after each route update to produce source rates and congeston prces for that update perod. These congeston prces may represent deays or oss probabtes across network nks. They determne the next routng update n the case of dynamc routng. Thus TCP AQM/IP form a feedback system where routng nteracts wth congeston contro n an teratve process. We are nterested n the equbrum and stabty propertes of ths teratve process. 1 Even though OSPF mpements a shortest-path agorthm, t aows mutpe equa-cost paths to be utzed. Our mode gnores ths feature. To smpfy notaton, we w henceforth use TCP AQM/IP and TCP/IP nterchangeaby. We assume routng s chosen to mnmze the weghted sum ap + bτ of congeston prces p and propagaton deays τ aong the path. When b = 0 pure dynamc routng, [26] characterzes the exact condton under whch an equbrum of TCP/IP exsts, and proves that such an equbrum maxmzes deay-nsenstve utty over both rates and routes. In practce, however, pure dynamc routng s never used because of ts nstabty. Instead, both weghts a and b are typcay nonzero, a case for whch no resut s avaabe. To reverse engneer TCP/IP networks wth nonzero weghts a and b, we ntroduce n Secton III deay-senstve utty functons that depend on not ony source rates but aso propagaton deays. We dentfy a cass C of deay-senstve utty functons that s mpcty optmzed by TCP/IP. As for the b = 0 case, we characterze the exact condton under whch TCP/IP has an equbrum and prove that such an equbrum maxmze utty functons n C over both rates and routes. As the reatve weght a/b, the utty functons n C become deay-nsenstve and these resuts reduce to those proved n [26] for pure dynamc routng. We exhbt some counter-ntutve propertes of cass C utty functons, and prove that any other cass of utty functons that TCP/IP optmzes necessary possess some strange propertes. We prove n Secton IV that, for genera networks, f the weght a s sma enough, ony mnmum-propagaton-deay paths are seected. Ths mpes that f a source-destnaton pars have unque mnmum-propagaton-deay paths, then equbrum of TCP/IP exsts and s gobay asymptotcay stabe. It s often beeved that decreasng a heps ensure routng stabty. We prove that ths may not be the case f not a source-destnaton pars have unque mnmum-propagatondeay paths. Indeed, for a genera network, ts equbrum and stabty propertes are the same as a modfed network whose routng s based on pure congeston prces p, a network that s prone to routng nstabty. More surprsngy, there exsts networks where reducng the weght a can destabze an orgnay stabe equbrum. It s conjectured n [26] that there s generay an nevtabe tradeoff between utty maxmzaton and stabty n TCP/IP networks. In partcuar, as the weght a ncreases, the routng s conjectured to become more unstabe but the achevabe utty hgher. We show however how to construct a network that has any gven utty profe as a functon of the weght a X/07/$ IEEE 418

2 II. MODEL We use the same mode as n [26]. In genera, we use sma etters to denote vectors, e.g., x wth x as ts th component; capta etters to denote matrces, e.g., H, W, R, or constants, e.g., L, N, K ; and scrpt etters to denote sets of vectors or matrces, e.g., W s, W m, R s, R m. Superscrpt s used to denote vectors, matrces, or constants pertanng to source, e.g., y, w, H, K. A. Network A network s modeed as a set of L undrectona nks shared by a set of N source-destnaton pars, ndexed by we w aso refer to the par smpy as source. Each nk has a fnte capacty c > 0 and a deay τ > 0 across the nk,.e., t takes τ to process and propagate a packet from one end of the nk to the other, excudng queueng deay. Let c =c, =1,...,L and τ =τ, =1,...,L. There are K acycc paths for source represented by a L K 0-1 matrx H where Hj 1, f path j of source uses nk = 0, otherwse Let H be the set of a coumns of H that represents a the avaabe paths to under snge-path routng. Defne the L K matrx H as H = [H 1... H N ] where K := K. H defnes the topoogy of the network. Let w be a K 1 vector where the jth entry represents the fracton of s fow on ts jth path such that w j 0 j and 1 T w =1 where 1 s a vector of an approprate dmenson wth the vaue 1 n every entry. We requre wj 0, 1} for sngepath routng, and aow wj [0, 1] for mut-path routng. Coect the vectors w, =1,...,N,ntoaK N bockdagona matrx W.LetW s be the set of a such matrces correspondng to snge path routng defned as W W = dagw 1,...,w N 0, 1} K N, 1 T w =1} Defne the correspondng set W m for mut-path routng as: W W = dagw 1,...,w N [0, 1] K N, 1 T w =1} As mentoned above, H defnes the set of acycc paths avaabe to each source, and represents the network topoogy. W defnes how the sources oad baance across these paths. Ther product defnes a L N routng matrx R = HW that specfes the fracton of s fow at each nk. Thesetofa snge-path routng matrces s R s = R R = HW,W W s } 1 and the set of a mut-path routng matrces s R m = R R = HW,W W m } 2 The dfference between snge-path routng and mut-path routng s the nteger constrant on W and R. A snge-path routng matrx n R s s an 0-1 matrx: 1, f nk s n a path of source R = 0, otherwse A mut-path routng matrx n R m s one whose entres are n the range [0, 1]: > 0, f nk s n a path of source R = 0, otherwse The path of source s denoted by r =[R 1... R L ] T, the th coumn of the routng matrx R. B. TCP AQM/IP We consder the stuaton where TCP AQM operates at a faster tmescae than routng updates. We assume a snge path s seected for each source-destnaton par that mnmzes the sum of the nk costs n the path, for some approprate defnton of nk cost. In partcuar, traffc s not spt across mutpe paths from the source to the destnaton even f they are avaabe. Ths modes, e.g., IP routng wthn an Autonomous System. We focus on the tmescae of the route changes, and assume TCP AQM s stabe and converges nstanty to equbrum after a route change. As n [17], we w nterpret the equbra of varous TCP and AQM agorthms as soutons of a utty maxmzaton probem defned n [12]. Dfferent TCP agorthms sove the same prototypca probem 3 wth dfferent utty functons; see e.g. [17], [19], [25] for the utty functons for varous popuar TCP proposas. Specfcay, suppose each source has a utty functon U x,d whch depends on both ts tota transmsson rate x and the end-to-end propagaton deay d. Gven a routng matrx R, we assume L d := d R = R τ Hence the deay d depends ony on routng R and not on congeston n the path. The routng matrx R s n R s for snge-path routng and n R m for mut-path routng. Note that n the mut-path case, d s the traffc-weghted average of propagaton deays aong ts paths. We assume that utty functons are strcty concave for fxed d. The speca case where the utty functon U x =U x,d depends on ts rate x but not on the deay d s studed n [26]. Here, we focus on the deay-senstve case. Gven a routng matrx R, U x,d s a functon ony of rate x.letrt R s be the snge-path routng n perod t. GvenaRt, et the equbrum rates xt = xrt and prces pt =prt generated by TCP AQM n perod t, respectvey, be the optma soutons of the constraned maxmzaton probem max x 0 U x,d s. t. Rtx c 3 and ts Lagrangan dua mn max U x,d x R tp + p 0 x 0 c p 4 419

3 The prces p t, =1,...,L, are measures of congeston, such as queueng deays or oss probabtes [17], [19]. We assume that the nk costs n perod t are z t = ap t+bτ 5 where a 0, b 0, and τ > 0 are constants. Based on these costs, each source computes ts new route r t +1 H ndvduay that mnmzes the sum of nk cost n ts path: r t + 1 = arg mn r H z tr 6 Reca that τ n 5 are propagaton deays across nks. If p t represents the queueng deays at nks and a = b =1, then z t represent tota deays across nks. The protoco parameters a and b determne the responsveness of routng to network traffc: a =0corresponds to statc routng, b = 0 corresponds to purey dynamc routng, and the arger the rato of a/b, the more responsve routng s to network traffc. They determne whether an equbrum exsts, whether t s stabe, and the achevabe utty at equbrum. The paper [26] focuses on the case of b =0; we study the genera case here. An equvaent way to specfy the TCP AQM/IP system as a dynamca system, at the tmescae of route changes, s to repace 3 4 by ther optmaty condtons. The routng s updated accordng to combnng 5 and 6 r t + 1 = arg mn r H ap t+bτ r, for a 7 where pt and xt are gven by [ ] + U R tp t = x t,d for a 8 c f p R tx t t 0 for a 9 = c f p t > 0 xt 0, pt 0 10 Ths set of equatons descrbe how the routng Rt, rates xt, and prces pt evove. Note that xt and pt depend on Rt ony through 8 10, mpcty assumng that TCP AQM converges nstanty to an equbrum gven the new routng Rt. We say that R,x,p s an equbrum of TCP/IP f t s a fxed pont of 3 6, or equvaenty, 7 10,.e., startng from routng R and assocated x,p, the above teratons yed R,x,p n the subsequent perods. C. The jont optmzaton probem Defnton 1: A deay-senstve utty functon s a contnuousy dfferentabe functon Ux, d from [0, [0, to [,, that satsfes the foowng propertes: 1 fxed d>0, Ux, d s strcty concave n x. 2 d >0,x>0, Ux, d and U x, d are fnte. 3 x > 0, d 1,d 2 s.t. 0 < d 1 < d 2 : Ux, d 1 > Ux, d 2. 4 D >0, 0 <d<d, Xd > 0, x <Xd : x, d > 0. U Essentay, a deay-senstve utty functon s defned so that the source aways gans utty from reducng propagaton deay. If propagaton deay s too hgh, the source can choose not to transmt. Otherwse, for fxed deay, the source s utty ncreases wth transmsson rate, possby up to some mt. We assume a sources on the network have deay-senstve utty functons U x,d. We adapt the snge-path deay-nsenstve network optmzaton probem from [26] to a deay-senstve network optmzaton probem: N L max U x, R τ s.t. Rx c 11 R R s,x 0 =1 Its Lagrangan dua s: N L mn max max L U x,d x R p + c p p 0 x 0 r =1 H 12 where r s the th coumn of R wth r = R. Ths probem maxmzes utty over both rates and routes. Defne the Lagrangan [1]: N L L LR, x, p = U x,d x R p + c p =1 Then we can express the prma and dua probems respectvey as: V sp = max max R R s V sd = mn max p 0 mn x 0 p 0 max R R s x 0 LR, x, p LR, x, p If we aow sources to use mutpe paths, the correspondng probems are: V mp = max max R R m V md = mn max p 0 mn x 0 p 0 max R R m x 0 LR, x, p LR, x, p The TCP/IP dynamca system s descrbed by 3 6, or equvaenty, D. Revew: deay-nsenstve utty functons In ths subsecton, we consder the speca case where the utty functons U x,d =U x depend ony on rates x but not on propagaton deays d. There are three sub-cases: a>0,b=0; a =0,b>0; a>0,b>0. For the frst case where a>0,b=0n 5,.e., IP uses ony congeston prces p generated by TCP AQM as nk costs, t s shown n [26] that TCP/IP maxmzes aggregate utty over both rates and routng when an equbrum exsts. Theorem 1 [26]: Suppose a>0 and b =0n 5. Then: 1 An equbrum R,x,p of TCP/IP exsts f and ony f there s no duaty gap between 11 and In ths case, the equbrum R,x,p s a souton of 11 and 12. Moreover, n that case, there s no penaty n not spttng the traffc among mutpe paths. Theorem 2 [26]: V sp V sd = V mp = V md. 420

4 One of the open questons rased n [26] s the characterzaton of TCP/IP equbrum n the other two cases where b>0 n 5,.e., when IP uses propagaton deay, excusvey or not, as nk costs. It s shown n [24] that for any deaynsenstve utty functon Ux, there exsts a network wth sources usng ths utty functon, where TCP/IP equbrum exsts but does not sove 11 and 12. See [24] for expct constructon of such networks. We now show that TCP/IP turns out to maxmze a cass of deay-senstve utty functons when b>0. III. DELAY-SENSITIVE NETWORK OPTIMIZATION In ths secton, we consder the case where the utty functons U x,d depend both on rates x and on propagaton deays d. We dentfy a cass C of utty functons for whch TCP/IP, wth a > 0 and b = 1 n 5, does maxmze aggregate utty at equbrum, when equbrum exsts. We anayze the propertes of C, and then derve propertes that any cass of utty functons that TCP/IP mpcty maxmzes at equbrum must possess. We start wth the case where a =0or b =0n 5,.e., f a nks use ony the propagaton deays τ as nk costs, or f a nks use ony the congeston prces p generated by TCP AQM as nk costs. In ths case, t can be shown that for every deay-senstve utty functon Ux, d, there exsts a network wth sources usng ths utty functon, where TCP/IP equbrum exsts but does not sove 11 and 12. See [24] for expct constructon of such networks. Hence we consder the case where both a>0 and b>0. A. Reverse engneerng for nk cost ap + τ In ths subsecton, we assume that IP uses ap + τ as nk cost,.e., a>0 and b =1n 5. Consder the cass C of functons Ux, d that can be wrtten as: Ux, d =V x a 1 xd where V x s a contnuousy dfferentabe functon from [0, to [, so that V x s strcty concave ncreasng, and x >0, V x and V x are fnte. In [24] we show that functons n C are deay-senstve utty functons defned n Defnton 1. Every utty functon n C has two components: V x measures the beneft of throughput x; a 1 xd measures the penaty of deay d weghted by the throughput x. Note that a arger throughput x ncreases both the beneft of throughput and the penaty due to deay. The reatve mportance s determned by the reatve weght a on congeston prce n the nk cost used n routng decsons: the arger the weght a, the more mportant the throughput beneft and the ess mportant the deay penaty. In the mtng case as a, correspondng to pure dynamc routng.e., b = 0, the utty functon become deay-nsenstve and our resuts beow reduce to those estabshed n [26] for the b =0case. Specazng to utty functons n C, the optmzaton probem 11 reduces to: N L max V x a 1 x R τ s.t. Rx c R R s,x 0 =1 Consder the Lagrangan LR, x, p = N L L V x x R p + a 1 τ + p c and the dua probem Dp := max R Rs,x 0 LR, x, p: Dp = N max x 0 =1 [V x x mn r H L L R p + a 1 τ ] + p c The mnmzaton over R n the dua probem appears to nvove mnma-cost routng usng p + a 1 τ as route cost. But ths s the same as mnma-cost routng usng ap + τ as route cost. Ths suggests that TCP/IP mght sove the jont optmzaton probem wth utty functons n C. Ths s ndeed the case. Our frst man resuts, and ther proofs, are anaogous to Theorem 1 and Theorem 2 but for b>0 and deay-senstve utty functons. They say that the equbrum of TCP/IP, when exsts, soves the jont utty maxmzaton over routes and rates and ts dua probem. Moreover, n that case, there s no penaty n not spttng traffc among mutpe paths. Ther proofs are n [24]. Theorem 3: Suppose a utty functons are n C, a>0 and b =1. 1 An equbrum R,x,p of TCP/IP exsts f and ony f there s no duaty gap between 11 and In ths case, the equbrum R,x,p s a souton of 11 and 12. Theorem 4: Suppose utty functons are n C. Then V sp V sd = V mp = V md. B. Counter-ntutve propertes of cass C In ths subsecton, we exhbt some counter-ntutve propertes of the cass C of utty functons. Specfcay, we present exampe networks where, because the penaty term a 1 xd s proportona to throughput x, these utty functons can underutze nk capactes or avaabe network paths. The frst exampe ustrates a network equbrum whch s n the strct nteror of the feasbe set Rx c, contrary to what the tradtona TCP mode wth deay-nsenstve utty functons woud predct. Remark 1: Gven any utty functon n C, there exsts a network where TCP/IP underutzes nks. Proof: U x, d =V x a 1 d.ifv x a 1 d =0, then x s the rate that maxmzes Ux, d for fxed d, snce Ux, d s strcty concave for fxed d. Choose any c > 0 and set τ = av c. Note that τ > 0, snce V x s strcty ncreasng. Consder a network wth one nk, whose capacty s 2c. A fow whose path s just ths nk w have rate c at equbrum, snce U x, d = V c a 1 τ = V c V c =0. But ths eaves the nk underutzed, snce the nk has capacty 2c. The second exampe shows that extra paths that woud be utzed f the utty functons were deay-nsenstve may not 421

5 Src Fg. 1. Network 1 L1: Capacty: c 1, Deay: τ 1 L2: Capacty: c 2, Deay: τ 2 Dest be utzed by utty functons n C. It aso ustrates Theorem 3 and Theorem 4. Consder any Ux, d = V x a 1 xd n C. It can be shown that there exst τ > 0, c>0 so that U c, τ > 0. Consder Network 1 n Fgure 1 wth c 1 = c, τ 1 = τ,c 2 =,τ 2 = τ + c, τ. Suppose there s ony one fow, Fow 1, and t can choose between routes R1: L1 and R2: L2. Suppose Fow 1 s ntay on route R1. It acheves rate c wth propagaton deay τ and utty Uc, τ. Then R1 has route cost τ + c, τ and R2 has route cost τ +a U c, τ. The nta routng s an equbrum routng snce a fows are usng mnma cost routes. Theorem 3 then mpes that there s no duaty gap. Theorem 4 then mpes that V sp = V mp and there s no beneft n mut-path routng,.e., there s no beneft n utzng route R2. Ths seems counter-ntutve wth a deay-nsenstve networks, there s aways a beneft n utzng prevousy unutzed routes. Indeed, we can show drecty that utzng R2 ncreases the average propagaton deay experenced by the fow, whch turns out to be suboptma regardess of the amount of extra throughput. Remark 2: Gven any utty functons n C, t s suboptma to use route R2 n Network 1 wth c 1 = c, τ 1 = τ,c 2 =,τ 2 = τ + c, τ, even when the fow s aowed to dstrbute ts traffc over mutpe paths. Proof: Let kc specfy the throughput on nk L2, so that k k+1 specfes the fracton of traffc sent over nk L2 where the rest s sent over nk L1. Then the tota throughput s gven by kc + c, and the weghted average propagaton deay s k [ τ + c, τ] + τ k +1 We cam that for every k>0, wemusthave Uc, τ U kc + c, k [ τ + c, τ] + τ k +1 To verfy ths, pug n our utty functon: V x a 1 cτ U kc + c, k τ + c, τ + τ k +1 V kc + c a 1 c kav c+τ V V kc + c V c c kc whch s true snce V x s concave. C. Aternatve casses We have shown n the ast subsecton some counter-ntutve propertes of cass C utty functons. We now show that any cass of deay-senstve utty functons that TCP/IP optmzes, usng ap + τ propertes. Defne as nk cost, must possess some strange MU, d := m Uc, d, 13 c whch computes the maxmum possbe utty at each deay d for any deay-senstve utty functon Ux, d. The next resut says that any cass B of utty functons that TCP/IP optmzes ether contan functons that are not strcty ncreasng n throughput contrary to the tradtona deay-nsenstve utty functons of TCP modes, or are dscontnuous n throughput n the space of utty functons, or can be dscontnuous n deay for arge enough deay. Theorem 5: Suppose B s any cass of deay-senstve utty functons such that when TCP/IP equbrum exsts, the equbrum soves 11 and 12 wth utty functons n B. Then B must have at east one of the foowng three propertes: 1 Ux, d B,d > 0 so that Ux, d s not strcty ncreasng n x. 2 U 1 x, d B, ɛ >0, wehaveu 2 x, d :=U 1 x + ɛ, d s not n B. 3 Ux, d B, D>0such that fd :=MU, d s fnte and dscontnuous for a d>d. Proof: See [24]. In partcuar, t can be checked that cass C utty functons possess the frst two propertes. IV. STABILITY AND UTILITY OF ROUTING POLICIES In ths secton, we anayze the effects on stabty and utty of dynamc routng. A. Suffcent condton for stabty Denote the set of paths G H wth mnma propagaton deay for fow by: } G := r H : τ T r = mn τ T s s H Denote the set of paths F H wthout mnma propagaton deay for fow by: F := H G Defne qr, a functon that computes the equbrum congeston prce vector for a gven routng matrx R R s. We assume that t mpcty depends on an arbtrary, fxed network L, N, F, G, H, R s,k,u,τ,c: N qr :=arg mn max U x,d p T Rx + p T c p 0 x 0 In ths subsecton, we show that wth suffcenty sma a,a fows w choose ony mnma propagaton deay paths. For notatona smpcty, a of the foowng functons, emmas, 422

6 and theorems n ths subsecton mpcty depend on an arbtrary, fxed network L, N, F, G, H, R s,k,u,τ,c. Defne hx, y, a functon that w be used to smpfy notaton: x y f y>0 hx, y = f y 0 Defne a # as foows: max mn bm,r a # := mn mn m G r F R R s 0< N f F f F = where bm,r :=h τ T r m,qr T m r. It can be checked that a # s strcty postve. Theorem 6: Suppose a<a #. Then R R s, 0 < N, m G such that aqr+τ T m < aqr+τ T r, r F 14 In other words, a fows on the network w choose a path wth mnma propagaton deay n the next teraton. Proof: We manpuate the rght hand sde of 14: aqr+τ T m < aqr+τ T r aqr T m r <τ T r m But by nspectng the defnton of hx, y, ths nequaty hods f a<h τ T r m,qr T m r snce τ T r m > 0 f m G and r F. The forma part of the theorem s then easy to see. It mpes that for any current routng, and for every fow, a path wth mnma propagaton deay has strcty ower cost than any path wthout mnma propagaton deay. Therefore no fow w seect any path wthout mnma propagaton deay. Therefore every fow w seect a path wth mnma propagaton deay. Theorem 7: Suppose a source-destnaton pars on a network have unque mnmum-propagaton-deay paths. Then f a<a #, TCP/IP has an asymptotcay stabe equbrum. Proof: Each fow ony has one path wth mnma propagaton deay. Appyng Theorem 6, each fow w aways seect the same path, and t w aways do ths from any routng. Coroary 1: Suppose every path n a network has dfferent propagaton deay. Then f a<a #, TCP/IP has an asymptotcay stabe equbrum. Proof: If every path n the network has dfferent propagaton deay, every source-destnaton par on a network has a unque mnmum-propagaton-deay path, so the resut of Theorem 7 appes. B. Counter-ntutve propertes: stabty It s often beeved that decreasng the weght a on the traffc senstve component of nk cost can aways stabze dynamc routng; e.g., see [3], [16], [26]. In ths subsecton, we show that ths s generay not true. It s easy to see that not a networks can be stabzed by decreasng a. For nstance, consder Network 1 wth c 1 = c 2 = c>0, τ 1 = τ 2 = τ>0. Suppose there s one fow, Fow 1, that can choose between routes R1: L1 and R2: L2. Ifthas a deay-nsenstve utty functon Ux, then for a a > 0, ths network has no equbrum. The next two resuts are ess obvous. The frst says that f a s sma enough, then a network wth routng based on ap + τ behaves ke a modfed network wth routng based on p, whch seems prone to routng nstabty [26]. Theorem 8: Gve a network wth a < a #. Consder the modfed network obtaned by deetng a paths wthout mnma propagaton deay from the orgna network. Then the orgna network wth routng based on ap + d has the same equbrum and stabty propertes as the modfed network wth routng based on p. Proof: Consder the TCP/IP dynamca system on the modfed network when nk costs are p : pt T r t + 1 = mn r G ptt r, for a pt,xt = arg mn max N U x,d p T Rtx+p T c p 0 x 0, t : r t G Consder the TCP/IP dynamca system on the orgna network when nk costs are ap + τ : apt+τ T r t + 1 = mn r H apt+τ T r, for a. pt,xt = arg mn max N U x,d p T Rtx+p T c p 0 x 0, t : r t H The next emma, whose proof s n [24], mpes that these dynamca systems are equvaent. Snce they are equvaent, they share the same equbrum and stabty propertes. Lemma 1: Suppose we have some network, and routng pocy on ths network s such that a<a #. Then for any attanabe prce vectors p so that p = qr for some R R s, and for a 0 < N, f and ony f p T r = mn s G pt s, where r G 15 ap + τ T r = mn s H ap + τ T s, where r H 16 The second counter-ntutve resut says that t s possbe to destabze a network by decreasng the statc component a n nk cost. Theorem 9: Consder any deay-senstve or deaynsenstve utty functon Ux, d. There exsts a network wth sources usng ths utty functon, and constants a 3 >a 2 a 1 > 0 so that: 1 The network s stabe for a 0,a

7 2 The network s unstabe for a a 2,a 3. 3 The network s stabe for a a 3,. We w prove the theorem by exhbtng such a network. Suppose Ux, d s an arbtrary deay-senstve or deaynsenstve functon. It can be shown [24] that t s possbe to choose parameters c 1,c 2,τ 1,τ 2 that satsfy the foowng nequates: U c 1,τ 1 > U c 2,τ 2 > 0 17 c 2 >c 1 > 0 18 τ 2 >τ 1 > 0 19 Consder Network 1 wth those parameters. Suppose there are two fows. Suppose the possbe routes are R1: L1 and R2: L2. Suppose that Fow 1 s constraned to R1, and Fow 2 can choose between R1 and R2. Denote ths nstance of Network 1byNetwork2. Defne functon A 3 U, c 1,τ 1,c 2,τ 2, where U s a deaynsenstve or deay-senstve utty functon and the rest of the parameters are n R, as foows: τ 2 τ 1 A 3 U, c 1,τ 1,c 2,τ 2 := U c 1,τ 1 U c 2,τ 2 Lemma 2: Consder any deay-senstve or deay-nsenstve utty functon Ux, d. Suppose c 1,τ 1,c 2,τ 2 satsfy 17 through 19 wth Ux, d. Network 2 wth a sources usng Ux, d s stabe for a a>a 3 U, c 1,τ 1,c 2,τ 2. Proof: We show that routng converges from every possbe nta condton. Suppose Fow 2 s on route R1. Then Fow 1 and Fow 2 share L1 equay, and both acheve rate c1 2 wth propagaton deay τ 1 and utty U c1 2,τ 1. Then R1 has route cost c1 2,τ 1+τ 1 and R2 has route cost τ 2. It can be verfed that a>a 3 U, c 1,τ 1,c 2,τ 2 mpes that: c1 1 2,τ + τ 1 >τ 2 So R2 s a ower cost route than R1. To show asymptotc stabty, t s then suffcent to show that the routng where Fow 2 s on R2 s an equbrum. Suppose Fow 2 s on R2. Then Fow 1 acheves rate c 1 wth propagaton deay τ 1 and utty Uc 1,τ 1, and Fow 2 acheves rate c 2 wth propagaton deay τ 2 and utty Uc 2,τ 2. Then R1 has route cost c 1,τ 1 +τ 1 and R2 has route cost c 2,τ 2 +τ 2. Consder Fow 2 s route choce at the next routng teraton. It can be verfed that a>a 3 U, c 1,τ 1,c 2,τ 2 mpes that c 1,τ 1 +τ 1 > c 2,τ 2 +τ 2 So R2 s a ower cost route than R1. Therefore R2 s mnma cost, and so ths routng s an equbrum routng. Defne functon A 2 U, c 1,τ 1,c 2,τ 2, where U s a deaynsenstve or deay-senstve utty functon and the rest of the parameters are n R as foows: A 2 U, c 1,τ 1,c 2,τ 2 := τ 2 τ 1 U c1 2,τ 1 Lemma 3: Consder any deay-senstve or deay-nsenstve utty functon Ux, d. Suppose c 1,τ 1,c 2,τ 2 satsfy 17 through 19 wth Ux, d. Then A 2 U, c 1,τ 1,c 2,τ 2 < A 3 U, c 1,τ 1,c 2,τ 2. Proof: See [24]. Lemma 4: Consder any deay-senstve or deay-nsenstve utty functon Ux, d. Suppose c 1,τ 1,c 2,τ 2 satsfy 17 through 19 wth Ux, d. Network 2 wth a sources usng Ux, d s unstabe for a a satsfyng A 2 U, c 1,τ 1,c 2,τ 2 < a<a 3 U, c 1,τ 1,c 2,τ 2. Proof: Suppose Fow 2 s on R2. Then Fow 1 acheves rate c 1 wth propagaton deay τ 1 and utty Uc 1,τ 1, and Fow 2 acheves rate c 2 wth propagaton deay τ 2 and utty Uc 2,τ 2. Then R1 has route cost c 1,τ 1 +τ 1 and R2 has route cost c 2,τ 2 +τ 2. Consder Fow 2 s route decson at the next routng teraton. It can be verfed that a<a 3 U, c 1,τ 1,c 2,τ 2 mpes c 1,τ 1 +τ 1 < c 2,τ 2 +τ 2. So Fow 2 next chooses route R1. Then Fow 1 and Fow 2 share L1 equay, and both acheve rate c1 2 wth deay τ 1 and utty U c1 2,τ 1. Then R1 has route cost c1 2,τ 1+τ 1 and R2 has route cost τ 2. It can be verfed that a>a 2 U, c 1,τ 1,c 2,τ 2 mpes c 1 2,τ 1+τ 1 >τ 2. So Fow 2 next chooses route R2. Ths mpes that Fow 2 s routng oscates between R1 and R2, so the network s unstabe. Proof Theorem 9. We choose c 1,τ 1,c 2,τ 2 so that they satsfy 17 through 19. Then consder Network 2 wth a sources usng Ux, d. Set a 1 = a # for ths network. Theorem 6 mpes that for a<a 1, the network s stabe. We then set a 2 = A 2 U, c 1,τ 1,c 2,τ 2 and a 3 = A 3 U, c 1,τ 1,c 2,τ 2.The emmas n ths subsecton then estabsh the desred resut. C. Counter-ntutve propertes: utty The paper [26] anayzes the effects of ncreasng a on the tme-averaged aggregate utty for a rng network, and a randomy generated network. On the rng network, tmeaveraged aggregate utty approached the maxmum possbe tme-averaged aggregate utty for any a, as a ncreased. On the generated network, tme-averaged aggregate utty ncreased unt routng stabty set n, and then decreased. In ths subsecton, we show that the effects of ncreasng a on tme-averaged aggregate utty are network dependent. In partcuar, we show how to construct a network wth any gven utty profe as a functon of the weght a. For every deay-nsenstve Ux, there exst k + >k > k > 0 and g > 0 so that g = Uk + Uk = Uk Uk. Aso, for every z [ 1, 1], there exsts k [k,k + ] so that Uk Uk =zg, that s gven by kz :=U 1 Uk +zg. Ths s easy to see snce Ux s strcty monotone ncreasng. 424

8 Consder the foowng network Nj, z, parameterzed by j>0, and z [ 1, 1], whch s defned to be Network 1 wth parameters c 1 = k,τ 1 = ju k,c 2 = kz,τ 2 =2jU k, wth one fow, Fow 1, choosng between routes R1: L1 and R2: L2. Denote the tme-averaged utty of the network under routng pocy ap + d by T N,a. Lemma 5: For every j>0, z [ 1, 1], network Nj, z has the foowng propertes: 1 a 1 [0,j],a 2 [0,j]: T N,a 1 =TN,a 2 2 a 1 j,,a 2 j, : T N,a 1 =TN,a 2 3 a 1 0,j],a 2 j, : T N,a 2 =TN,a 1 + zg 2 Proof: Suppose current routng s R2. Then R1 has route cost ju k and R2 has route cost 2jU k +au kz. By nspecton, for every a 0, R1 has ess cost than R2. Therefore routng on the next teraton w be R1. Suppose current routng s R1. Then R1 has route cost ju k +au k and R2 has route cost 2jU k. By nspecton, f a j then R1 s a ower cost route than R2, and f a>j, then R2 s a strcty ower cost route than R1. Fow 1 s utty on route R1 between routng changes s Uk. Fow 1 s utty on route R2 between routng changes s Ukz. If a j, then the network s stabe wth Fow 1 on R1 and the aggregate utty s Uk.Ifa > j, then the network oscates between R1 and R2 and acheves tmeaveraged utty Ukz + Uk /2. The excess utty ganed by ncreasng a from ess than j to more than j s gven by: Ukz + Uk 2 Uk = Ukz Uk 2 = zg 2 Defnton 2: A utty-versus-a profe s a par of vectors x, y such that x = y > 0, : x > 0, and < x : x <x +1. Defnton 3: A network N matches a utty-versus-a profe x, y f there exsts λ>0 so that: If x > 1, 1 s.t. 1 < < x, a 1 s.t. x 1 < a 1 x, a 2 s.t. x <a 2 <x +1 : T N,a 2 T N,a 1 =λy. 2 a 1 s.t. 0 < a 1 x 1, a 2 s.t. x 1 < a 2 < x 2 : T N,a 2 T N,a 1 =λy 1. 3 a 1 s.t. x x 1 <a 1 x x, a 2 s.t. x x <a 2 < : T N,a 2 T N,a 1 =λy x. If x =1, a 1 s.t. 0 <a 1 x 1, a 2 s.t. x 1 <a 2 < : T N,a 2 T N,a 1 =λy 1. In other words, for a, the tme-averaged aggregate utty of the network ncreases by λy at a = x. Theorem 10: For every utty-versus-a profe x, y, there exsts a network wth sources usng deay-nsenstve utty functons that matches ths profe. Proof: Consder any utty-versus-a profe x, y. Defne the normazed y as y := max y 1 y. Construct the network N by takng the unon of networks N 1...N y where, N := Nx,y. The subnetworks are entrey dsjont n the unon network. It s easy to see that the network N matches profe x, y g wth λ = 2 max v. V. CONCLUSION In ths paper, we have attempted to reverse engneer TCP/IPke networks. We have dentfed a cass of deay-senstve utty functons that s mpcty optmzed by an equbrum of TCP/IP. We have characterzed ts equbrum and stabty propertes for genera networks, and exhbted severa counterntutve resuts. Many ssues are st open. Frst, we beeve C s the ony cass of utty functons that TCP/IP wth nk cost ap + d jonty optmze, but we have not been abe to prove ths. Second, the deay-nsenstve utty functons obtaned from reverse engneerng TCP agorthms n the terature, assumng routng s fxed, are a strcty ncreasng n throughput. Hence the deay-senstve utty functons obtaned from reverse engneerng TCP/IP shoud deay be strcty ncreasng n throughput when routng s hed fxed. However, ths s not the case wth cass C utty functons. Ths msmatch shoud be reconced. REFERENCES [1] D. Bertsekas. Nonnear Programmng. Athena Scentfc, [2] D. Bertsekas, E. Gafn, and R. G. Gaager. Second dervatve agorthms for mnmum deay dstrbuted routng n networks. IEEE Trans. on Communcatons, 328: , [3] Dmtr P. Bertsekas. Dynamc behavor of shortest path routng agorthms for communcaton networks. IEEE Transactons on Automatc Contro, pages 60 74, February [4] S. Deb and R. Srkant. Congeston contro for far resource aocaton n networks wth mutcast fows. In Proceedngs of the IEEE Conference on Decson and Contro, December [5] R. G. Gaager. A mnmum deay routng agorthm usng dstrbuted computaton. IEEE Trans. on Communcatons, 251:73 85, [6] R. G. Gaager and S. J. Goestan. Fow contro and routng agorthms for data networks. In Proc. of the 5th Internatona Conf. Comp. Comm., pages , [7] Jayue He, Mung Chang, and Jennfer Rexford. TCP/IP nteracton based on congeston prce: Stabty and optmaty. In Proc. IEEE ICC, June [8] C. Hedrck. Routng Informaton Protoco. IETF RFC 1058, http: // June [9] K. Kar, S. Sarkar, and L. Tassuas. Optmzaton based rate contro for mutpath sessons. In Proceedngs of 7th Internatona Teetraffc Congress ITC, December [10] Koushk Kar, Saswat Sarkar, and Leandros Tassuas. Optmzaton based rate contro for mutrate mutcast sessons. In Proceedngs of IEEE Infocom, Apr [11] F. P. Key and T. Voce. Stabty of end-to-end agorthms for jont routng and rate contro. In ACM Sgcomm Computer Communcaton Revew, voume 35, pages 5 12, Apr [12] Frank P. Key, Aman Mauoo, and Davd Tan. Rate contro for communcaton networks: Shadow prces, proportona farness and stabty. Journa of Operatons Research Socety, 493: , [13] S. Kunnyur and R. Srkant. End-to-end congeston contro: utty functons, random osses and ECN marks. IEEE/ACM Transactons on Networkng, 115: , October [14] I. Lestas and G. Vnncombe. Combned contro of routng and fow: a mutpath routng approach. In Proc. 43rd IEEE Conference on Decson and Contro, December [15] X. Ln and N. B. Shroff. Utty maxmzaton for communcaton networks wth mut-path routng. IEEE Trans. on Automatc Contro, 515: , May [16] S. Low and P. Varaya. Dynamc behavor of a cass of adaptve routng protocos IGRP. In Proceedngs of Infocom 93, pages , March

9 [17] Steven H. Low. A duaty mode of tcp and queue management agorthms. IEEE/ACM Trans. Netw., 114: , [18] Steven H. Low and Davd E. Lapsey. Optmzaton fow contro, I: basc agorthm and convergence. IEEE/ACM Transactons on Networkng, 76: , December [19] Steven H. Low, Larry L. Peterson, and Lmn Wang. Understandng tcp vegas: a duaty mode. J. ACM, 492: , [20] L. Massoue and J. Roberts. Bandwdth sharng: objectves and agorthms. IEEE/ACM Transactons on Networkng, 103: , June [21] Jeonghoon Mo and Jean Warand. Far end-to-end wndow-based congeston contro. IEEE/ACM Transactons on Networkng, 85: , October [22] J. Moy. OSPF Verson 2. IETF RFC 2328, org/rfcs/rfc2328.htm, Apr [23] F. Pagann. Congeston contro wth adaptve mutpath routng based on optmzaton. In Proc. of CISS, March [24] John Pongsajapan. Optmzaton and stabty of tcp/p wth deaysenstve utty functons. Master s thess, Caforna Insttute of Technoogy, [25] R. Srkant. The Mathematcs of Internet Congeston Contro. Brkhauser, [26] Jantao Wang, Lun L, Steven H. Low, and John C. Doye. Cross-ayer optmzaton n tcp/p networks. IEEE/ACM Trans. Netw., 133: ,

Optimization Models for Heterogeneous Protocols

Optimization Models for Heterogeneous Protocols Optmzaton Modes for Heterogeneous Protocos Steven Low CS, EE netab.caltech.edu wth J. Doye, S. Hegde, L. L, A. Tang, J. Wang, Catech M. Chang, Prnceton Outne Internet protocos Horzonta decomposton TCP-AQM

More information

Optimization and Stability of TCP/IP with Delay-Sensitive Utility Functions

Optimization and Stability of TCP/IP with Delay-Sensitive Utility Functions Optimization and Stability of TCP/IP with Delay-Sensitive Utility Functions Thesis by John Pongsajapan In Partial Fulfillment of the Requirements for the Degree of Master of Science California Institute

More information

Research on Complex Networks Control Based on Fuzzy Integral Sliding Theory

Research on Complex Networks Control Based on Fuzzy Integral Sliding Theory Advanced Scence and Technoogy Letters Vo.83 (ISA 205), pp.60-65 http://dx.do.org/0.4257/ast.205.83.2 Research on Compex etworks Contro Based on Fuzzy Integra Sdng Theory Dongsheng Yang, Bngqng L, 2, He

More information

Associative Memories

Associative Memories Assocatve Memores We consder now modes for unsupervsed earnng probems, caed auto-assocaton probems. Assocaton s the task of mappng patterns to patterns. In an assocatve memory the stmuus of an ncompete

More information

A finite difference method for heat equation in the unbounded domain

A finite difference method for heat equation in the unbounded domain Internatona Conerence on Advanced ectronc Scence and Technoogy (AST 6) A nte derence method or heat equaton n the unbounded doman a Quan Zheng and Xn Zhao Coege o Scence North Chna nversty o Technoogy

More information

MARKOV CHAIN AND HIDDEN MARKOV MODEL

MARKOV CHAIN AND HIDDEN MARKOV MODEL MARKOV CHAIN AND HIDDEN MARKOV MODEL JIAN ZHANG JIANZHAN@STAT.PURDUE.EDU Markov chan and hdden Markov mode are probaby the smpest modes whch can be used to mode sequenta data,.e. data sampes whch are not

More information

Pricing and Resource Allocation Game Theoretic Models

Pricing and Resource Allocation Game Theoretic Models Prcng and Resource Allocaton Game Theoretc Models Zhy Huang Changbn Lu Q Zhang Computer and Informaton Scence December 8, 2009 Z. Huang, C. Lu, and Q. Zhang (CIS) Game Theoretc Models December 8, 2009

More information

Interference Alignment and Degrees of Freedom Region of Cellular Sigma Channel

Interference Alignment and Degrees of Freedom Region of Cellular Sigma Channel 2011 IEEE Internatona Symposum on Informaton Theory Proceedngs Interference Agnment and Degrees of Freedom Regon of Ceuar Sgma Channe Huaru Yn 1 Le Ke 2 Zhengdao Wang 2 1 WINLAB Dept of EEIS Unv. of Sc.

More information

Xin Li Department of Information Systems, College of Business, City University of Hong Kong, Hong Kong, CHINA

Xin Li Department of Information Systems, College of Business, City University of Hong Kong, Hong Kong, CHINA RESEARCH ARTICLE MOELING FIXE OS BETTING FOR FUTURE EVENT PREICTION Weyun Chen eartment of Educatona Informaton Technoogy, Facuty of Educaton, East Chna Norma Unversty, Shangha, CHINA {weyun.chen@qq.com}

More information

Neural network-based athletics performance prediction optimization model applied research

Neural network-based athletics performance prediction optimization model applied research Avaabe onne www.jocpr.com Journa of Chemca and Pharmaceutca Research, 04, 6(6):8-5 Research Artce ISSN : 0975-784 CODEN(USA) : JCPRC5 Neura networ-based athetcs performance predcton optmzaton mode apped

More information

ON AUTOMATIC CONTINUITY OF DERIVATIONS FOR BANACH ALGEBRAS WITH INVOLUTION

ON AUTOMATIC CONTINUITY OF DERIVATIONS FOR BANACH ALGEBRAS WITH INVOLUTION European Journa of Mathematcs and Computer Scence Vo. No. 1, 2017 ON AUTOMATC CONTNUTY OF DERVATONS FOR BANACH ALGEBRAS WTH NVOLUTON Mohamed BELAM & Youssef T DL MATC Laboratory Hassan Unversty MORO CCO

More information

A General Column Generation Algorithm Applied to System Reliability Optimization Problems

A General Column Generation Algorithm Applied to System Reliability Optimization Problems A Genera Coumn Generaton Agorthm Apped to System Reabty Optmzaton Probems Lea Za, Davd W. Cot, Department of Industra and Systems Engneerng, Rutgers Unversty, Pscataway, J 08854, USA Abstract A genera

More information

Deriving the Dual. Prof. Bennett Math of Data Science 1/13/06

Deriving the Dual. Prof. Bennett Math of Data Science 1/13/06 Dervng the Dua Prof. Bennett Math of Data Scence /3/06 Outne Ntty Grtty for SVM Revew Rdge Regresson LS-SVM=KRR Dua Dervaton Bas Issue Summary Ntty Grtty Need Dua of w, b, z w 2 2 mn st. ( x w ) = C z

More information

ECE559VV Project Report

ECE559VV Project Report ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate

More information

A parametric Linear Programming Model Describing Bandwidth Sharing Policies for ABR Traffic

A parametric Linear Programming Model Describing Bandwidth Sharing Policies for ABR Traffic parametrc Lnear Programmng Mode Descrbng Bandwdth Sharng Poces for BR Traffc I. Moschoos, M. Logothets and G. Kokknaks Wre ommuncatons Laboratory, Dept. of Eectrca & omputer Engneerng, Unversty of Patras,

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

Note 2. Ling fong Li. 1 Klein Gordon Equation Probablity interpretation Solutions to Klein-Gordon Equation... 2

Note 2. Ling fong Li. 1 Klein Gordon Equation Probablity interpretation Solutions to Klein-Gordon Equation... 2 Note 2 Lng fong L Contents Ken Gordon Equaton. Probabty nterpretaton......................................2 Soutons to Ken-Gordon Equaton............................... 2 2 Drac Equaton 3 2. Probabty nterpretaton.....................................

More information

Complete subgraphs in multipartite graphs

Complete subgraphs in multipartite graphs Complete subgraphs n multpartte graphs FLORIAN PFENDER Unverstät Rostock, Insttut für Mathematk D-18057 Rostock, Germany Floran.Pfender@un-rostock.de Abstract Turán s Theorem states that every graph G

More information

Achieving Optimal Throughput Utility and Low Delay with CSMA-like Algorithms: A Virtual Multi-Channel Approach

Achieving Optimal Throughput Utility and Low Delay with CSMA-like Algorithms: A Virtual Multi-Channel Approach Achevng Optma Throughput Utty and Low Deay wth SMA-ke Agorthms: A Vrtua Mut-hanne Approach Po-Ka Huang, Student Member, IEEE, and Xaojun Ln, Senor Member, IEEE Abstract SMA agorthms have recenty receved

More information

Lower bounds for the Crossing Number of the Cartesian Product of a Vertex-transitive Graph with a Cycle

Lower bounds for the Crossing Number of the Cartesian Product of a Vertex-transitive Graph with a Cycle Lower bounds for the Crossng Number of the Cartesan Product of a Vertex-transtve Graph wth a Cyce Junho Won MIT-PRIMES December 4, 013 Abstract. The mnmum number of crossngs for a drawngs of a gven graph

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

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

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

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

Lower Bounding Procedures for the Single Allocation Hub Location Problem

Lower Bounding Procedures for the Single Allocation Hub Location Problem Lower Boundng Procedures for the Snge Aocaton Hub Locaton Probem Borzou Rostam 1,2 Chrstoph Buchhem 1,4 Fautät für Mathemat, TU Dortmund, Germany J. Faban Meer 1,3 Uwe Causen 1 Insttute of Transport Logstcs,

More information

Price Competition under Linear Demand and Finite Inventories: Contraction and Approximate Equilibria

Price Competition under Linear Demand and Finite Inventories: Contraction and Approximate Equilibria Prce Competton under Lnear Demand and Fnte Inventores: Contracton and Approxmate Equbra Jayang Gao, Krshnamurthy Iyer, Huseyn Topaogu 1 Abstract We consder a compettve prcng probem where there are mutpe

More information

Research Article H Estimates for Discrete-Time Markovian Jump Linear Systems

Research Article H Estimates for Discrete-Time Markovian Jump Linear Systems Mathematca Probems n Engneerng Voume 213 Artce ID 945342 7 pages http://dxdoorg/11155/213/945342 Research Artce H Estmates for Dscrete-Tme Markovan Jump Lnear Systems Marco H Terra 1 Gdson Jesus 2 and

More information

Delay tomography for large scale networks

Delay tomography for large scale networks Deay tomography for arge scae networks MENG-FU SHIH ALFRED O. HERO III Communcatons and Sgna Processng Laboratory Eectrca Engneerng and Computer Scence Department Unversty of Mchgan, 30 Bea. Ave., Ann

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

Subgradient Methods and Consensus Algorithms for Solving Convex Optimization Problems

Subgradient Methods and Consensus Algorithms for Solving Convex Optimization Problems Proceedngs of the 47th IEEE Conference on Decson and Contro Cancun, Mexco, Dec. 9-11, 2008 Subgradent Methods and Consensus Agorthms for Sovng Convex Optmzaton Probems Björn Johansson, Tamás Kevczy, Mae

More information

Polite Water-filling for Weighted Sum-rate Maximization in MIMO B-MAC Networks under. Multiple Linear Constraints

Polite Water-filling for Weighted Sum-rate Maximization in MIMO B-MAC Networks under. Multiple Linear Constraints 2011 IEEE Internatona Symposum on Informaton Theory Proceedngs Pote Water-fng for Weghted Sum-rate Maxmzaton n MIMO B-MAC Networks under Mutpe near Constrants An u 1, Youjan u 2, Vncent K. N. au 3, Hage

More information

A General Distributed Dual Coordinate Optimization Framework for Regularized Loss Minimization

A General Distributed Dual Coordinate Optimization Framework for Regularized Loss Minimization Journa of Machne Learnng Research 18 17 1-5 Submtted 9/16; Revsed 1/17; Pubshed 1/17 A Genera Dstrbuted Dua Coordnate Optmzaton Framework for Reguarzed Loss Mnmzaton Shun Zheng Insttute for Interdscpnary

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

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

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton

More information

Assortment Optimization under MNL

Assortment Optimization under MNL Assortment Optmzaton under MNL Haotan Song Aprl 30, 2017 1 Introducton The assortment optmzaton problem ams to fnd the revenue-maxmzng assortment of products to offer when the prces of products are fxed.

More information

Analysis of Bipartite Graph Codes on the Binary Erasure Channel

Analysis of Bipartite Graph Codes on the Binary Erasure Channel Anayss of Bpartte Graph Codes on the Bnary Erasure Channe Arya Mazumdar Department of ECE Unversty of Maryand, Coege Par ema: arya@umdedu Abstract We derve densty evouton equatons for codes on bpartte

More information

Cyclic Codes BCH Codes

Cyclic Codes BCH Codes Cycc Codes BCH Codes Gaos Feds GF m A Gaos fed of m eements can be obtaned usng the symbos 0,, á, and the eements beng 0,, á, á, á 3 m,... so that fed F* s cosed under mutpcaton wth m eements. The operator

More information

Lecture 21: Numerical methods for pricing American type derivatives

Lecture 21: Numerical methods for pricing American type derivatives Lecture 21: Numercal methods for prcng Amercan type dervatves Xaoguang Wang STAT 598W Aprl 10th, 2014 (STAT 598W) Lecture 21 1 / 26 Outlne 1 Fnte Dfference Method Explct Method Penalty Method (STAT 598W)

More information

Outline. Communication. Bellman Ford Algorithm. Bellman Ford Example. Bellman Ford Shortest Path [1]

Outline. Communication. Bellman Ford Algorithm. Bellman Ford Example. Bellman Ford Shortest Path [1] DYNAMIC SHORTEST PATH SEARCH AND SYNCHRONIZED TASK SWITCHING Jay Wagenpfel, Adran Trachte 2 Outlne Shortest Communcaton Path Searchng Bellmann Ford algorthm Algorthm for dynamc case Modfcatons to our algorthm

More information

Single-Source/Sink Network Error Correction Is as Hard as Multiple-Unicast

Single-Source/Sink Network Error Correction Is as Hard as Multiple-Unicast Snge-Source/Snk Network Error Correcton Is as Hard as Mutpe-Uncast Wentao Huang and Tracey Ho Department of Eectrca Engneerng Caforna Insttute of Technoogy Pasadena, CA {whuang,tho}@catech.edu Mchae Langberg

More information

This article reviews the current transmission. Internet Congestion Control. By Steven H. Low, Fernando Paganini, and John C. Doyle.

This article reviews the current transmission. Internet Congestion Control. By Steven H. Low, Fernando Paganini, and John C. Doyle. DIGITAL VISION LTD. Internet Congeston Contro By Steven H. Low Fernando Pagann and John C. Doye Ths artce revews the current transmsson contro protoco (TCP congeston contro protocos and overvews recent

More information

Appendix for An Efficient Ascending-Bid Auction for Multiple Objects: Comment For Online Publication

Appendix for An Efficient Ascending-Bid Auction for Multiple Objects: Comment For Online Publication Appendx for An Effcent Ascendng-Bd Aucton for Mutpe Objects: Comment For Onne Pubcaton Norak Okamoto The foowng counterexampe shows that sncere bddng by a bdders s not aways an ex post perfect equbrum

More information

Structure and Drive Paul A. Jensen Copyright July 20, 2003

Structure and Drive Paul A. Jensen Copyright July 20, 2003 Structure and Drve Paul A. Jensen Copyrght July 20, 2003 A system s made up of several operatons wth flow passng between them. The structure of the system descrbes the flow paths from nputs to outputs.

More information

Achieving Optimal Throughput Utility and Low Delay with CSMA-like Algorithms: A Virtual Multi-Channel Approach

Achieving Optimal Throughput Utility and Low Delay with CSMA-like Algorithms: A Virtual Multi-Channel Approach IEEE/AM TRANSATIONS ON NETWORKING, VOL. X, NO. XX, XXXXXXX 20X Achevng Optma Throughput Utty and Low Deay wth SMA-ke Agorthms: A Vrtua Mut-hanne Approach Po-Ka Huang, Student Member, IEEE, and Xaojun Ln,

More information

Distributed Moving Horizon State Estimation of Nonlinear Systems. Jing Zhang

Distributed Moving Horizon State Estimation of Nonlinear Systems. Jing Zhang Dstrbuted Movng Horzon State Estmaton of Nonnear Systems by Jng Zhang A thess submtted n parta fufment of the requrements for the degree of Master of Scence n Chemca Engneerng Department of Chemca and

More information

NP-Completeness : Proofs

NP-Completeness : Proofs NP-Completeness : Proofs Proof Methods A method to show a decson problem Π NP-complete s as follows. (1) Show Π NP. (2) Choose an NP-complete problem Π. (3) Show Π Π. A method to show an optmzaton problem

More information

Queueing Networks II Network Performance

Queueing Networks II Network Performance Queueng Networks II Network Performance Davd Tpper Assocate Professor Graduate Telecommuncatons and Networkng Program Unversty of Pttsburgh Sldes 6 Networks of Queues Many communcaton systems must be modeled

More information

Nested case-control and case-cohort studies

Nested case-control and case-cohort studies Outne: Nested case-contro and case-cohort studes Ørnuf Borgan Department of Mathematcs Unversty of Oso NORBIS course Unversty of Oso 4-8 December 217 1 Radaton and breast cancer data Nested case contro

More information

Singular Value Decomposition: Theory and Applications

Singular Value Decomposition: Theory and Applications Sngular Value Decomposton: Theory and Applcatons Danel Khashab Sprng 2015 Last Update: March 2, 2015 1 Introducton A = UDV where columns of U and V are orthonormal and matrx D s dagonal wth postve real

More information

A MIN-MAX REGRET ROBUST OPTIMIZATION APPROACH FOR LARGE SCALE FULL FACTORIAL SCENARIO DESIGN OF DATA UNCERTAINTY

A MIN-MAX REGRET ROBUST OPTIMIZATION APPROACH FOR LARGE SCALE FULL FACTORIAL SCENARIO DESIGN OF DATA UNCERTAINTY A MIN-MAX REGRET ROBST OPTIMIZATION APPROACH FOR ARGE SCAE F FACTORIA SCENARIO DESIGN OF DATA NCERTAINTY Travat Assavapokee Department of Industra Engneerng, nversty of Houston, Houston, Texas 7704-4008,

More information

A A Non-Constructible Equilibrium 1

A A Non-Constructible Equilibrium 1 A A Non-Contructbe Equbrum 1 The eampe depct a eparabe contet wth three payer and one prze of common vaue 1 (o v ( ) =1 c ( )). I contruct an equbrum (C, G, G) of the contet, n whch payer 1 bet-repone

More information

L-Edge Chromatic Number Of A Graph

L-Edge Chromatic Number Of A Graph IJISET - Internatona Journa of Innovatve Scence Engneerng & Technoogy Vo. 3 Issue 3 March 06. ISSN 348 7968 L-Edge Chromatc Number Of A Graph Dr.R.B.Gnana Joth Assocate Professor of Mathematcs V.V.Vannaperuma

More information

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal Inner Product Defnton 1 () A Eucldean space s a fnte-dmensonal vector space over the reals R, wth an nner product,. Defnton 2 (Inner Product) An nner product, on a real vector space X s a symmetrc, blnear,

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

LOW-DENSITY Parity-Check (LDPC) codes have received

LOW-DENSITY Parity-Check (LDPC) codes have received IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 59, NO. 7, JULY 2011 1807 Successve Maxmzaton for Systematc Desgn of Unversay Capacty Approachng Rate-Compatbe Sequences of LDPC Code Ensembes over Bnary-Input

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

On the Power Function of the Likelihood Ratio Test for MANOVA

On the Power Function of the Likelihood Ratio Test for MANOVA Journa of Mutvarate Anayss 8, 416 41 (00) do:10.1006/jmva.001.036 On the Power Functon of the Lkehood Rato Test for MANOVA Dua Kumar Bhaumk Unversty of South Aabama and Unversty of Inos at Chcago and Sanat

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

An Augmented Lagrangian Coordination-Decomposition Algorithm for Solving Distributed Non-Convex Programs

An Augmented Lagrangian Coordination-Decomposition Algorithm for Solving Distributed Non-Convex Programs An Augmented Lagrangan Coordnaton-Decomposton Agorthm for Sovng Dstrbuted Non-Convex Programs Jean-Hubert Hours and Con N. Jones Abstract A nove augmented Lagrangan method for sovng non-convex programs

More information

find (x): given element x, return the canonical element of the set containing x;

find (x): given element x, return the canonical element of the set containing x; COS 43 Sprng, 009 Dsjont Set Unon Problem: Mantan a collecton of dsjont sets. Two operatons: fnd the set contanng a gven element; unte two sets nto one (destructvely). Approach: Canoncal element method:

More information

Lecture 4. Instructor: Haipeng Luo

Lecture 4. Instructor: Haipeng Luo Lecture 4 Instructor: Hapeng Luo In the followng lectures, we focus on the expert problem and study more adaptve algorthms. Although Hedge s proven to be worst-case optmal, one may wonder how well t would

More information

How Bad Is Suboptimal Rate Allocation?

How Bad Is Suboptimal Rate Allocation? How Bad Is Suboptmal Rate Allocaton? Tan Lan, Xaojun Ln 2, Mung Chang, Ruby Lee Department of Electrcal Engneerng, Prnceton Unversty, NJ 08544, USA 2 School of Electrcal and Computer Engneerng, Purdue

More information

A Simple Congestion-Aware Algorithm for Load Balancing in Datacenter Networks

A Simple Congestion-Aware Algorithm for Load Balancing in Datacenter Networks IEEE/ACM TRANSACTIONS ON NETWORKING A Smpe Congeston-Aware Agorthm for Load Baancng n Datacenter Networs Mehrnoosh Shafee, Student Member, IEEE, and Javad Ghader, Member, IEEE Abstract We study the probem

More information

Greyworld White Balancing with Low Computation Cost for On- Board Video Capturing

Greyworld White Balancing with Low Computation Cost for On- Board Video Capturing reyword Whte aancng wth Low Computaton Cost for On- oard Vdeo Capturng Peng Wu Yuxn Zoe) Lu Hewett-Packard Laboratores Hewett-Packard Co. Pao Ato CA 94304 USA Abstract Whte baancng s a process commony

More information

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan Wnter 2008 CS567 Stochastc Lnear/Integer Programmng Guest Lecturer: Xu, Huan Class 2: More Modelng Examples 1 Capacty Expanson Capacty expanson models optmal choces of the tmng and levels of nvestments

More information

Dynamic Systems on Graphs

Dynamic Systems on Graphs Prepared by F.L. Lews Updated: Saturday, February 06, 200 Dynamc Systems on Graphs Control Graphs and Consensus A network s a set of nodes that collaborates to acheve what each cannot acheve alone. A network,

More information

Optimal Guaranteed Cost Control of Linear Uncertain Systems with Input Constraints

Optimal Guaranteed Cost Control of Linear Uncertain Systems with Input Constraints Internatona Journa Optma of Contro, Guaranteed Automaton, Cost Contro and Systems, of Lnear vo Uncertan 3, no Systems 3, pp 397-4, wth Input September Constrants 5 397 Optma Guaranteed Cost Contro of Lnear

More information

Downlink Power Allocation for CoMP-NOMA in Multi-Cell Networks

Downlink Power Allocation for CoMP-NOMA in Multi-Cell Networks Downn Power Aocaton for CoMP-NOMA n Mut-Ce Networs Md Shpon A, Eram Hossan, Arafat A-Dwe, and Dong In Km arxv:80.0498v [eess.sp] 6 Dec 207 Abstract Ths wor consders the probem of dynamc power aocaton n

More information

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud

Resource Allocation with a Budget Constraint for Computing Independent Tasks in the Cloud Resource Allocaton wth a Budget Constrant for Computng Independent Tasks n the Cloud Wemng Sh and Bo Hong School of Electrcal and Computer Engneerng Georga Insttute of Technology, USA 2nd IEEE Internatonal

More information

e - c o m p a n i o n

e - c o m p a n i o n OPERATIONS RESEARCH http://dxdoorg/0287/opre007ec e - c o m p a n o n ONLY AVAILABLE IN ELECTRONIC FORM 202 INFORMS Electronc Companon Generalzed Quantty Competton for Multple Products and Loss of Effcency

More information

Formulas for the Determinant

Formulas for the Determinant page 224 224 CHAPTER 3 Determnants e t te t e 2t 38 A = e t 2te t e 2t e t te t 2e 2t 39 If 123 A = 345, 456 compute the matrx product A adj(a) What can you conclude about det(a)? For Problems 40 43, use

More information

Monica Purcaru and Nicoleta Aldea. Abstract

Monica Purcaru and Nicoleta Aldea. Abstract FILOMAT (Nš) 16 (22), 7 17 GENERAL CONFORMAL ALMOST SYMPLECTIC N-LINEAR CONNECTIONS IN THE BUNDLE OF ACCELERATIONS Monca Purcaru and Ncoeta Adea Abstract The am of ths paper 1 s to fnd the transformaton

More information

The Second Anti-Mathima on Game Theory

The Second Anti-Mathima on Game Theory The Second Ant-Mathma on Game Theory Ath. Kehagas December 1 2006 1 Introducton In ths note we wll examne the noton of game equlbrum for three types of games 1. 2-player 2-acton zero-sum games 2. 2-player

More information

Lecture 10 Support Vector Machines II

Lecture 10 Support Vector Machines II Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed

More information

Solutions to exam in SF1811 Optimization, Jan 14, 2015

Solutions to exam in SF1811 Optimization, Jan 14, 2015 Solutons to exam n SF8 Optmzaton, Jan 4, 25 3 3 O------O -4 \ / \ / The network: \/ where all lnks go from left to rght. /\ / \ / \ 6 O------O -5 2 4.(a) Let x = ( x 3, x 4, x 23, x 24 ) T, where the varable

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

606 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 27, NO. 5, JUNE Abdallah Khreishah, Chih-Chun Wang, and Ness B.

606 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 27, NO. 5, JUNE Abdallah Khreishah, Chih-Chun Wang, and Ness B. 66 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL 7, NO 5, JUNE 9 Cross-Layer Optmzaton for Wreess Muthop Networks wth Parwse Intersesson Network Codng Abdaah Khreshah, Chh-Chun Wang, and Ness B

More information

Image Classification Using EM And JE algorithms

Image Classification Using EM And JE algorithms Machne earnng project report Fa, 2 Xaojn Sh, jennfer@soe Image Cassfcaton Usng EM And JE agorthms Xaojn Sh Department of Computer Engneerng, Unversty of Caforna, Santa Cruz, CA, 9564 jennfer@soe.ucsc.edu

More information

Supplementary Notes for Chapter 9 Mixture Thermodynamics

Supplementary Notes for Chapter 9 Mixture Thermodynamics Supplementary Notes for Chapter 9 Mxture Thermodynamcs Key ponts Nne major topcs of Chapter 9 are revewed below: 1. Notaton and operatonal equatons for mxtures 2. PVTN EOSs for mxtures 3. General effects

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

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

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper Games of Threats Elon Kohlberg Abraham Neyman Workng Paper 18-023 Games of Threats Elon Kohlberg Harvard Busness School Abraham Neyman The Hebrew Unversty of Jerusalem Workng Paper 18-023 Copyrght 2017

More information

A Class of Distributed Optimization Methods with Event-Triggered Communication

A Class of Distributed Optimization Methods with Event-Triggered Communication A Cass of Dstrbuted Optmzaton Methods wth Event-Trggered Communcaton Martn C. Mene Mchae Ubrch Sebastan Abrecht the date of recept and acceptance shoud be nserted ater Abstract We present a cass of methods

More information

ON THE BEHAVIOR OF THE CONJUGATE-GRADIENT METHOD ON ILL-CONDITIONED PROBLEMS

ON THE BEHAVIOR OF THE CONJUGATE-GRADIENT METHOD ON ILL-CONDITIONED PROBLEMS ON THE BEHAVIOR OF THE CONJUGATE-GRADIENT METHOD ON I-CONDITIONED PROBEM Anders FORGREN Technca Report TRITA-MAT-006-O Department of Mathematcs Roya Insttute of Technoogy January 006 Abstract We study

More information

Integrating advanced demand models within the framework of mixed integer linear problems: A Lagrangian relaxation method for the uncapacitated

Integrating advanced demand models within the framework of mixed integer linear problems: A Lagrangian relaxation method for the uncapacitated Integratng advanced demand modes wthn the framework of mxed nteger near probems: A Lagrangan reaxaton method for the uncapactated case Mertxe Pacheco Paneque Shad Sharf Azadeh Mche Berare Bernard Gendron

More information

Random Walks on Digraphs

Random Walks on Digraphs Random Walks on Dgraphs J. J. P. Veerman October 23, 27 Introducton Let V = {, n} be a vertex set and S a non-negatve row-stochastc matrx (.e. rows sum to ). V and S defne a dgraph G = G(V, S) and a drected

More information

corresponding to those of Heegaard diagrams by the band moves

corresponding to those of Heegaard diagrams by the band moves Agebra transformatons of the fundamenta groups correspondng to those of Heegaard dagrams by the band moves By Shun HORIGUCHI Abstract. Ths paper gves the basc resut of [1](1997),.e., a hande sdng and a

More information

Numerical integration in more dimensions part 2. Remo Minero

Numerical integration in more dimensions part 2. Remo Minero Numerca ntegraton n more dmensons part Remo Mnero Outne The roe of a mappng functon n mutdmensona ntegraton Gauss approach n more dmensons and quadrature rues Crtca anass of acceptabt of a gven quadrature

More information

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin

LOW BIAS INTEGRATED PATH ESTIMATORS. James M. Calvin Proceedngs of the 007 Wnter Smulaton Conference S G Henderson, B Bller, M-H Hseh, J Shortle, J D Tew, and R R Barton, eds LOW BIAS INTEGRATED PATH ESTIMATORS James M Calvn Department of Computer Scence

More information

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION

CHAPTER-5 INFORMATION MEASURE OF FUZZY MATRIX AND FUZZY BINARY RELATION CAPTER- INFORMATION MEASURE OF FUZZY MATRI AN FUZZY BINARY RELATION Introducton The basc concept of the fuzz matr theor s ver smple and can be appled to socal and natural stuatons A branch of fuzz matr

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

Lecture 12: Discrete Laplacian

Lecture 12: Discrete Laplacian Lecture 12: Dscrete Laplacan Scrbe: Tanye Lu Our goal s to come up wth a dscrete verson of Laplacan operator for trangulated surfaces, so that we can use t n practce to solve related problems We are mostly

More information

Errors for Linear Systems

Errors for Linear Systems Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch

More information

Can Shortest-path Routing and TCP Maximize Utility

Can Shortest-path Routing and TCP Maximize Utility Can Shotest-path Routng and TCP Maxmze Utty Jantao Wang Lun L Steven H. Low John C. Doye Cafona Insttute of Technoogy, Pasadena, USA {jantao@cds., un@cds., sow@, doye@cds.}catech.edu Abstact TCP-AQM potoco

More information

Key words. corner singularities, energy-corrected finite element methods, optimal convergence rates, pollution effect, re-entrant corners

Key words. corner singularities, energy-corrected finite element methods, optimal convergence rates, pollution effect, re-entrant corners NESTED NEWTON STRATEGIES FOR ENERGY-CORRECTED FINITE ELEMENT METHODS U. RÜDE1, C. WALUGA 2, AND B. WOHLMUTH 2 Abstract. Energy-corrected fnte eement methods provde an attractve technque to dea wth eptc

More information

COS 521: Advanced Algorithms Game Theory and Linear Programming

COS 521: Advanced Algorithms Game Theory and Linear Programming COS 521: Advanced Algorthms Game Theory and Lnear Programmng Moses Charkar February 27, 2013 In these notes, we ntroduce some basc concepts n game theory and lnear programmng (LP). We show a connecton

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

LECTURE 21 Mohr s Method for Calculation of General Displacements. 1 The Reciprocal Theorem

LECTURE 21 Mohr s Method for Calculation of General Displacements. 1 The Reciprocal Theorem V. DEMENKO MECHANICS OF MATERIALS 05 LECTURE Mohr s Method for Cacuaton of Genera Dspacements The Recproca Theorem The recproca theorem s one of the genera theorems of strength of materas. It foows drect

More information

A Local Variational Problem of Second Order for a Class of Optimal Control Problems with Nonsmooth Objective Function

A Local Variational Problem of Second Order for a Class of Optimal Control Problems with Nonsmooth Objective Function A Local Varatonal Problem of Second Order for a Class of Optmal Control Problems wth Nonsmooth Objectve Functon Alexander P. Afanasev Insttute for Informaton Transmsson Problems, Russan Academy of Scences,

More information