Achieving High Throughput and Low Delay by Accurately Regulating Link Queue Length over Mobile Data Network

Size: px
Start display at page:

Download "Achieving High Throughput and Low Delay by Accurately Regulating Link Queue Length over Mobile Data Network"

Transcription

1 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) Achevng Hgh Throughpu and Low Delay by Accuraely Regulang Lnk ueue Lengh over Moble Daa ework e Lu Sae ey Laboraory o Compuer Archecure Insue o Compung Technology Chnese Academy o Scences Emal: luke@c.ac.cn Jack Y. B. Lee Deparmen o Inormaon Engneerng The Chnese Unversy o Hong ong, Hong ong Emal: yblee@e.cuhk.edu.hk Absrac nowledge o he queue lengh o he boleneck lnk has many poenal applcaons such as congeson conrol, rac engneerng, rac polcng, conen adapaon, os monorng and provsonng, ec. A recen work proposed a new Sum-o-Delay wh Tmesamp (SoD-TS) algorhm whch can accuraely esmae queue lengh n moble neworks wh boh bandwdh varaons and uplnk delay varaons by explong he exsng TCP Tmesamp opon. By makng use o SoD-TS we developed a novel ranspor proocol queue-lengh-aware TCP (TCP-LA), ha ackles he problem o buerbloa o provde beer os or applcaons requrng low end-o-end delay and/or hgh bandwdh ulzaon. Trace-drven smulaon shows ha compared o TCP CUBIC TCP LA can reduce he RTT by a acor o 2.7 whle sll achevng over 97% bandwdh ulzaon. Moreover, TCP-LA urher reduces RTT by 5% compared o delay-based TCP such as FAST TCP and TCP Vegas. Index Terms: queue lengh, TCP, bandwdh ulzaon. I. ITRODUCTIO TCP [-5] have snce evolved over he years and been adoped as he ranspor proocol or numerous Inerne applcaons. Gven he lack o os conrol n he curren Inerne s essenal o desgn TCP o adap o varaons n nework condons. Wh he emergence o moble daa neworks such as 3G/4G [6], he challenge o changng nework condons s even more crcal or applcaon desgners, conen provders as well as moble operaors. Among he TCP varans, TCP CUBIC [5], he domnaed TCP varan n use n Inerne servers, can acheve he good bandwdh ulzaon bu resuls n he problem o buerbloa [7] over moble daa neworks, as keeps ncreasng CWnd unl lnk buer overlow, hus buldng up a large backlog n he lnk buer. I s a ac ha queue lengh o boleneck lnk can have sgncan mpac o TCP s perormance. Thus one can accuraely esmae he queue lengh and regulae o a arge queue lengh, hen can manan low end-o-end delay whle sll achevng hgh bandwdh ulzaon. Prevous works such as Vegas mehod [2] and Sum-o-delays mehod (SoD) [8] only ncorporae he lnk bandwdh varaons n esmang he queue lengh bu do no consder delay varaons n he uplnk,.e., rom he moble devce o he Inerne, and smply assumed o be consan. Inerne server Inerne T w L U q c T m Moble daa neworks(e.g., 3G.LTE) Moble recever Fg. The sysem model or analyzng he TCP-LA algorhm. In mos Inerne applcaons he drecon o daa ranser s ypcally rom he Inerne o he moble devce. Thus racs n he uplnk are usually much less and prmarly consss o applcaon requess (e.g., HTTP reques) and AC packes generaed by he TCP recever. Hence uplnk bandwdh s rarely he boleneck. However, he recen measuremens [8, 9] conduced n producon 3G/LTE moble neworks revealed ha he uplnk delay can and do exhb some varaons, even s no congesed. We conjecure ha he varaons are due o acors such as lnk-layer reransmssons [6], rado resource schedulng, and queueng a he uplnk. aurally hs wll aec he esmaon accuraces o Vegas and SoD mehods. Our prevous work proposed a praccal soluon Sum-o-Delay wh Tmesamp (SoD-TS) [], o hs problem by explong TCP s Tmesamp Opon [] whch s wdely mplemened by moble devces. We urher showed ha SoD-TS can be appled o TCP CUBIC by explong he esmaed queue lengh as sgnals or TCP CUBIC s congeson wndow ncrease/decrease. In hs work we ake queue-lengh-based congeson conrol urher by developng a novel ranspor proocol queue-lengh-aware TCP (TCP-LA), o ackle he problem o buerbloa over moble daa neworks whle mananng hgh nework ulzaon. TCP-LA s desgned o regulae he queue lengh a he boleneck lnk o a arge queue lengh o manan low end-o-end delay whle sll achevng hgh bandwdh ulzaon. Boh sysem model and race-drven smulaons show ha compared o TCP CUBIC TCP-LA can reduce he round-rp me (RTT) by a acor o over 2.7 whle sll achevng over 97% bandwdh ulzaon. Addonally, compared o delay-based TCP such as FAST TCP and TCP Vegas, TCP-LA reduces RTT by 5% as /4/$3. 24 IEEE 562

2 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) requres a smaller arge queue lengh o acheve near % bandwdh ulzaon, hus always resuls n he lowes RTT. TCP-LA s parcularly desrable or applcaons requrng low end-o-end delay and hgh bandwdh ulzaon, such as lve vdeo sreamng. The res o he paper s organzed as ollows: Secon II revews some relaed works on queue lengh esmaon mehod. Secon III presens he proposed TCP-LA; Secon IV presens a sysem model o TCP-LA and proves s several neresng properes. Secon V evaluaes TCP-LA va race-drven smulaons. Secon VII summarzes he sudy and oulnes some uure work. II. UEUE LEGTH ESTIATIO METHOD In hs secon we brely revs several queue lengh esmaon mehods. A. Vegas Mehod Frs, some TCP varans do employ relaed echnques o eher mplcly or explcly esmae he queue lengh a he boleneck lnk or congeson conrol purpose. One well known example s TCP Vegas [2] whch employ an algorhm o esmae he queue lengh rom he congeson wndow sze and he derences beween curren and mnmum RTTs. Smlar algorhms have also been adoped n TCP Veno [4] and FAST TCP [3]. Speccally, he TCP sender connuously measures he RTT, denoed by r, and records he congeson wndow sze, denoed by cwnd, a he me AC packe s receved. I hen keeps rack o he mnmum r mn by, rmn mn r () and hen compue he esmaed queue lengh, denoed by n, rom he derence beween he expeced hroughpu,.e., cwnd r mn, and he acual measured hroughpu,.e., cwnd r, n he las round-rp me [2], ' cwnd( r rmn ) n (2) r Vegas mehod uses he congeson wndow sze o he curren RTT o esmae he queue lengh n he prevous RTT, whle FAST TCP addressed hs shorcomng by keepng rack o he pas congeson wndows o he prevous RTTs, and usng he one a he me o he orgnal daa ransmsson nsan o compue (2) or beer accuracy. Chan e.al [8] also showed ha he FAST mehod has he esmaon error o 4.8%, whle Vegas mehod has he esmaon error o a leas 5.8%. These esmaon errors are due o her nably o rack shor-me-scale bandwdh varaons common n moble daa neworks. Addonally, Vegas and FAST do no compensae or uplnk delay varaons and smply lump ogeher wh queueng delay varaons. Ths may lead o alse posves n congeson esmaon, resulng n lower bandwdh ulzaon. B. Sum-o-Delay Mehod To ackle he naccurae queue lengh esmaon due o bandwdh varaons Chan e.al nvened a new sum-o-delay (SoD) mehod whch compensaes or he mpac o nework bandwdh varaons n perormng queue lengh esmaons by drecly measurng he bandwdh rom AC packe mngs and keepng rack o he seres o pas bandwdh measuremens. The ollowng s a smpled descrpon o he SoD algorhm based on he sudy by Chan and Lee [8]. Fg. depcs he sysem model where a sngle sender ransms packes va he rado nerace o a recever. Le T w be he orwardng delay rom he Inerne server o he rado nerace; T m be he orwardng delay rom he rado nerace o he moble recever. The recever acknowledges receved packe by sendng back an AC packe o he sender whch akes u seconds or AC o reurn o he sender. In he orgnal SoD algorhm he uplnk delay s consan,.e., u =U or all. one o he boleneck lnk s parameers are known a pror. The goal s o esmae he queue lengh, denoed by q. The prncple o SoD s o make use o he seres o AC packe mngs o oban an esmae o he packe ransmsson me, he laer o whch ncorporaes he mpac o nework delay and bandwdh varaons. By dervng wo nequales,.e., (7) and (), whch bound he queueng delay, one could hen oban an esmae o he queue lengh n (2). Speccally, le be he ransmsson me or packe whch can be esmaed rom he measured AC packe ner-arrval mes or packes and (), denoed by r and r - respecvely: r r (3) provded ha he lnk does no become dle durng hs me. Suppose packe arrves a he boleneck lnk o nd ha here are n packes already queueng, plus packe ( n ) currenly beng ransmed. Then he queueng delay o packe, denoed by q, s equal o he oal me o ransm all he n queueng packes,.e., he sum o ransmsson delays o packe ( ) o ( n ), plus he resdual me o ransm packe ( n ), denoed by : q 2... n (4) where he resdual ransmsson me s bounded by n (5) Subsung (5) no (4) gves he correspondng bounds or q : 2... n q 2... n n (6) whch becomes he rs nequaly or q : k q k k n k n (7) ueue lengh esmaon s hen equvalen o ndng n. The only remanng unknown erm s q whch canno be measured drecly by he sender. However one can oban a lower bound or q. Consder he RTT or packe : r q U Tw Tm (8) and he mnmum RTT measured so ar: rmn mn{ q } U Tw Tm (9) Subrac (9) rom (8) elmnaes he unknown erms excep q : r rmn q mn{ q } () q q, Rearrangng we hen have he second nequaly or q : 563

3 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) q r r () mn wh all hree erms n he R.H.S. known o he sender. Fnally, we noe ha q mus sasy boh nequales n (7) and (), and hence we can esmae n rom: n mn x k r rmn, x,,... (2) k x The smalles n ha sases (2) hen becomes he esmaed queue lengh n. Ther resuls showed ha SoD can acheve sgncanly hgher esmaon accuracy even n moble neworks wh rapdly changng bandwdh. everheless, one assumpon n he prevous analyss s ha he uplnk delay s consan,.e., u =U,. Le λ, λ max λ, denoe he uplnk delay varaon experenced by he h AC packe receved by he sender. As delay varaons are relave we dene such ha he AC packe experencng he mnmum uplnk delay known as he reerence AC, has a delay varaon o zero. Comparng o he AC packe recepon me whou he delay varaon, denoed by r, he acual recepon me, denoed by r, now becomes r =r +λ (3) and he acual RTT measured, denoed by r, s gven by r =r +λ. (4) To ncorporae uplnk delay varaons we replace, r, and k n (2) wh, r, and k respecvely o oban ' ' ' n mn x k r rmn, x,,... (5) k x By rearrangng he erms (5) becomes n mn x k r rmn x2, x,,... (6) k x Comparng o (2) we can clearly see he mpac o uplnk delay varaon,.e., x 2, whch resul n over-esmaon o he queue lengh and s esmaon error s shown o be bounded by, max / mn packes. C. Sum-o-Delay wh Tmesamp Mehod To compensae he uplnk delay varaons we proposed a Sum-o-Delay wh Tmesamp (SoD-TS) mehod [], o urher mprove SoD s esmaon accuracy n neworks wh large uplnk delay varaons by explorng he TCP mesamp opon [] or hs propose. Speccally, assumng he recever generaes an AC mmedaely upon recevng a daa packe. Then he TCP sender can compue he daa packe ransmsson me drecly rom he correspondng AC packe s ransmsson mesamp TSval. Le r denoe he TSval o he h AC. Then n (6) can be compued rom =r r -. (7) ex we updae (5) by replacng he wh n (7), and subsung (4) no (5) o oban n mn x k r rmn, x,,... (8) k x Thus he only remanng unknown s. We observe ha aer he SoD algorhm converges, r mn would be known and we would also know whch packe,.e., he reerence packe, resuled n he mnmum RTT. Le he reerence packe be packe number re, and le r re and r be he arrval me and AC generaon me o he reerence packe respecvely. Le r denoe he acual AC recepon me and r denoe he AC generaon me. The uplnk delay varaon experenced by he h AC can be compued rom ( r ' rre ) ( r rre ) (9) We also showed ha he delay-varaon compensaed SoD-TS algorhm has a queue lengh esmaon error bounded by - packe,.e., underesmae by a mos one packe. Resuls rom race-drven smulaons also showed ha (a) ypcal uplnk delay varaons such as ms can degrade he esmaon accuraces o SoD by as much as %; and (b) he SoD-TS can eecvely elmnae he esmaon errors due o uplnk delay varaons, e.g.,.95%.esmaon error. III. UEUE-LEGTH-AWARE COGESTIO COTROL Armed wh he accurae knowledge o queue lengh rom SoD-TS we propose a novel ranspor proocol ueue-lengh-aware TCP (TCP-LA) ha connuously esmaes he queue lengh usng SoD-TS algorhm, hen deermnes he CWnd accordngly o accuraely regulae he queue lengh oscllang around a arge queue lengh, whch s able o manan low end-o-end delay whle sll achevng hgh bandwdh ulzaon. TCP-LA s organzed n wo key sages: () sar-up phase and () adapve phase. Speccally, le W denoe he nal CWnd, TCP-LA use a burs o W packes o provde an nal esmae o he BDP o he nework. Thereaer, TCP-LA adjuss he ransmsson rae o regulae he queue lengh oscllang a arge queue lengh usng SoD-TS algorhm. A. Sarup Phase I s a ac ha he modern moble neworks mplemen large rado lnk buers (e.g., 277.5B or 3G/HSPA). Moreover, he rado lnk buers are allocaed on a per user bass [2, 3] and hus are no shared even when mulple users are acve nsde he same moble cell. Compared o he large lnk buer sze, exsng TCP mplemenaons are oen oo conservave n resrcng he number o packes nlgh va he congeson wndow, hereby lmng he achevable hroughpu. Insead, TCP-LA sars a new connecon wh a burs o W packes. The exac value o W can be derved by explong knowledge o he properes o he moble nework. Ideally, s desrable or TCP-LA o ll he nework o more han he maxmum BDP. Speccally, le C max denoe he maxmum bandwdh o he rado nework (e.g., 7.2 Mbps or 3G/HSPA), and D max denoe he maxmum propagaon delay o he moble daa nework ncludng he maxmum uplnk delay varaons (e.g., 287ms or 3G/HSPA), hus TCP-LA wll se W =C max D max (e.g., 8 packes), s relavely small compared o he sze o he rado lnk buer (e.g., 852 packes), hus wll no lkely cause buer overlow. In conras, re 564

4 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) convenonal TCP adops a small nal congeson wndow sze n TCP (e.g., 2 o 4 packes). Aer recevng he rs AC packe rom he moble recever TCP-LA leaves he sarup phase and swches no he adapve phase. B. Adapve Phase As soon as recevng he s AC packe rom he moble recever TCP-LA eners he adapve phase over whch he subsequen ransmsson wll hen adap o he queue lengh esmaed va SoD-TS connuously. Smlar o TCP Vegas and FAST TCP, TCP-LA updaes CWnd accordng o esmaed queue lengh once per RTT. Speccally, le T denoe he arge queue lengh and n j denoe he esmaed queue lengh a he begnnng o he j h RTT. Upon recevng he rs AC o he j h RTT, n j T, TCP-LA ncreases CWnd by T n j, hus wll ransm T- n j more packes besdes one ransmed packe due o AC clockng, oherwse, TCP-LA decreases CWnd by n j T, hus wll sop ransmng unl recevng n j T ACs ncludng he curren receved AC. In summary, TCP-LA updaes CWnd accordng o he ollowng rule: le W j denoe he CWnd durng he j h RTT. Upon recevng he rs AC o he j h RTT we have W W n (2) j j T j C. Loss Recovery Unlke convenonal TCP varans, he CWnd o TCP-LA s no aeced by packe loss evens. oneheless, los packes sll need o be reransmed and A-TCP sll regulaes he queue lengh around T usng (2) by explong TCP SAC [4]. As A-TCP lms he amoun o packes n he boleneck lnk buer, mplcly avods he congeson loss snce he arge queue lengh s much less han he lnk buer sze. Addonally, A-TCP guaranees he ecen bandwdh ulzaon durng loss recovery, whle convenonal TCP varans, e.g., TCP CUBIC, TCP Reno, cu down her CWnd blndly whou consderng he bandwdh ulzaon. IV. MODELLIG AD AALYSIS In hs secon we develop a sysem model or TCP-LA ha can be easly exend o model delay-based TCP such as FAST TCP; esablsh s sucen condon o ull bandwdh ulzaon and derve he properes o TCP-LA. A. Background and Assumpons Fg. depcs he sysem model ha he Inerne server s ransmng packes o he moble recever. Our ocus s on he ecen ulzaon o he rado nework and hus we assume he Inerne server and he backhaul lnkng o he moble nework are no he boleneck. Speccally, One TCP connecon s esablshed when he moble recever aemps o esablsh a connecon o he Inerne server. Upon recep o daa reques he Inerne server sars o ransm he packes and we assume ha he Inerne server always has daa wang o be ransmed o he recever. Symbol T w Table. Summary o symbols and noaons. Descrpon Downlnk delay rom Inerne server o he rado nerace. Downlnk delay rom he rado nerace o he moble T m recever. Uplnk delay rom he moble recever o he Inerne server. U U, U s a random varable whle U s he delay or he rs AC. Round rp propagaon delay,.e, D Tm U Tw. D s a D, D random varable whle D s he delay or he rs AC. The uplnk delay varaon a he me, λ() λ λ() max, where λ max s he maxmum uplnk varaons The congeson wndow a he h RTT, W W s he nal congeson wndow The esmaed queue lengh esmaed a he h RTT n n The acual queue lengh a he h RTT σ The queue lengh esmaon error n erms o packe The me aer he h RTT The ransmsson me or he rs daa packe. The amoun o daa acknowledged by delayed ACs a me, () () λ max C max C mn Mnmum lnk capacy, C mn. C max Maxmum lnk capacy, C max C mn. () Lnk buer lengh a me (n uns o bye) The number o packes receved by he rado lnk buer rom n me o me The number o packes sen ou by he lnk buer rom me o ou me The number o packes receved by he TCP recever as A n repored by ACs receved by he Inerne server rom me o me A ou The number o packes sen ou by he Inerne server rom me o me The Inerne server s conneced o he rado nerace va hgh-speed (e.g., Gbps) xed nework lnks. Le U be he uplnk delay rom he moble recever o he server. Thus he round rp propagaon delay D s gven by D=T m +U+T w. We gnore delay varaons n he downlnk as has he same eec as bandwdh varaon,.e., aecng he ransmsson me and hus s accouned or by he bandwdh esmaon process. For he uplnk we allow he delay o vary bu s bounded wh he maxmum delay. Le λ() denoe he oal uplnk delay varaon a me and λ() λ max, where λ max s he maxmum delay. Hence we have U [ Umn, Umax ] and D [ Dmn, Dmax ], where U max =U mn +λ max and D max =D mn +λmax. For he rado nerace we assume he rado lnk bandwdh over an arbrary nerval, s bounded rom above by C max and rom below by C mn, where C max >C mn. Oher han hese bounds he rado bandwdh can vary n arbrary manner rom one nerval o he nex. Table summarzes he noaons and symbols used n he sysem model. B. Sysem Model To derve he queue lengh a he rado nerace we consder he amoun o daa arrvng a and deparng rom he queue a any gven me. To smply he analyss we employ he packe-based model or packe low n he ollowng dervaons. Frs, he queue lengh s smply equal o he derence beween ncomng and ougong packes: () (2) n ou 565

5 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) Where n are he number o packes arrved a an queue n he nerval (, ] and ou are he number o packes capable o be ransmed by he boleneck lnk n he nerval (, ]. oe ha () he bandwdh s ully ulzed and he value o () s queue lengh a me. ow consder he packe low rom he Inerne server o he queue. We noe ha he number o packes receved by he queue, e.g., n, s equal o he number o packes sen by he Inerne server T w seconds earler (c.., Fg. ): S Tw (22) n ou Tw S Tw where ou s he number o packes ransmed by he Tw Inerne server n he nerval ( T w, T w ]. As depced n Fg., le U be he uplnk delay experenced by he rs AC. Then he rs AC wll arrve a he server a me D +, where D =T m +U +T w. Durng he perod beore he rs AC arrves,.e., (, D + ], he ncomng daa rae o he queue s smply equal o he server s ransmsson rae n he perod (T w, D + T w ], hus D S D Tw W (23) n ou Tw n We dvde no wo sub-nervals (, D + ] and (D +, ] as ollows: D () n n D ou (24) Smlar o (23) we can oban he second erm n (24) rom ong ha Thus S Tw n D ou D Tw (25) D T ( T U T ) T w m w w Tm U S Tw n D ou T m U (26) (27) ex we consder he las erm n (24) by sub-dvdng no wo nervals (, D mn ] and (D mn, ] where D mn s he mnmum RTT: Dmn ou ou ou D (28) mn Subsung back no (24) we have D Dmn () n n D ou ou D (29) mn The rs erm on he R.H.S. o (28) s he amoun o daa sen by he rado nerace durng (, D mn ]. The ACs or hese daa are guaraneed o reurn back o he server whn he nerval [T m +U +, T w +λ max ] as delay varaon s bounded by λ max. Consder he ACs generaed rom packes ransmed by he rado lnk rom (, D mn ]. Le be he amoun o daa acknowledged by delayed ACs,.e., ACs arrvng durng [ T w, T w +λ max ] he server. We can hen rewre he rs erm n (28) as ollows: Dmn S Tw ou n (3) Tm U Where s bounded by maxcmax (3) Subsung (27) and (29) no (3) we hen have D S Tw S Tw () n ou n ou T D m U Tm U mn (32) By rearrangng he erms we have D S Tw S Tw () n ou D mn ou Tm U n Tm U (33) ex we consder he me nerval [ Tm U, Tw ] or he las wo erms n (33). Le M denoe he oal ACs receved a he nerne server durng [T m +U +, T w ], hus S Tw n M. Upon recevng every AC, TCP-LA has Tm U hree condons: ) TCP-LA ransms one daa packe due o TCP AC-clockng. 2) TCP-LA ransms T n more daa packes besdes one packe due o condon he receved AC s he rs AC o he h RTT and n T ; 3) TCP-LA s possble no o ransm any daa packe, snce TCP-LA sops ransmng unl recevng n T ACs n T when recevng he rs AC o he h RTT; Le m denoe he number o receved ACs ha TCP-LA goes o condon, hus M m s he number o receved ACs ha TCP-LA goes o condon 3. Le denoe he number o RTT durng [T m +U +, T w ]. Accordng o condon 2 and 3 TCP-LA wll ransm max( T n, ) packes upon recevng he rs AC o every RTT. Thereore we have, S Tw S Tw max n, mm ou Tm U n T Tm U (34) where m s he number o packes ransmed due o condon. Le k={x x } denoe he se o RTTs durng he nerval [T m +U +, T w ], le k denoe he subse o k ha condon 3 occurs, hus we have, T n M m (35) k We can om he max uncon n he rs erm o R.H.S. o (34) by addng he L.H.S. o (35) o (34), hus we have, S Tw S Tw n, n mm ou Tm U n T T Tm U k (36) Cancel he las wo erms o (36) accordng o (35) and Subsue (36) no (33) we have, () W ou D mn T n (37) Le n denoe he acual queue lengh, hus n =n +σ, where σ s he esmaon error resuled rom he queue lengh esmaon mehod a he begnnng o he h RTT. Hence (37) becomes, () W ou D mn T n (38) Rearrange he erms we have, 566

6 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) ou D mn T () W n (39) Le σ max dene he maxmum esmaon error resuled rom he queue lengh esmaon mehod and hus σ σ max or all. Subsue σ wh σ max we have he mnmum o (), mn(()) as ollows, () mn W n ou Dmn T max (4) Accordng o he above sysem model we can derve several neresng properes n he ollowng secons. C. Guaraneeng Full Bandwdh Ulzaon The sucen condon o ull bandwdh ulzaon s mn(()) or all, hus we have he ollowng heorem: THEOREM : One possble sucen condon o sasy mn(()) or all s W CmaxDmax (4) and T max n / (42) Proo: Obvously, subsue (4) and (42) no (4) we have () mn CmaxDmax ou D (43) mn As ou D + D mn mncmax maxcmax, we have () or all. oe ha he mnmum requred arge queue lengh s acheved when he equaly n (42) s sased. D. Targe ueue Lengh As shown n (42), TCP-LA needs o ncorporae σ max o se T o have he ull bandwdh ulzaon. In [] we proved ha SoD-TS s esmaon error s only bounded by [, ] packe, hus σ max, sod-s =. Smlarly, we also showed ha σ max, sod s bounded by, max / mn packes and σ max, as max / mn snce σ max, as mgh conan he esmaon error due o bandwdh varaons besdes he one due o uplnk delay varaons. I s obvous ha σ max, as >σ max, sod >σ max, sod-s, hus we have he ollowng heorem: THEOREM 2: Le T, qla, T, sod and T, as denoe he mnmum requred arge queue lengh o TCP-LA, TCP-SoD and FAST TCP o ully ulze he bandwdh under all nework condons, hus we have T, qla < T, sod < T, as. Proo: As he hree TCP varans updaes CWnd every RTT, we have derved rom (39) a r and =n +, where n + s he accurae esmaon o a he sender aer a eedback delay. Subsue wh n + n (39) we have + n = W ou D mn T (44) Smlarly, Subsue - wh n we have - = n W ou D mn T (45) Le (44) subrac (45) we have n ou Dmn ou Dmn T (46) Snce n + =, we have n + he proocol ully ulzes he bandwdh, hus we have T + ou + D mn ou D mn (47) Snce TCP-LA, TCP-SoD and FAST TCP experences he same nework condons, he las our erms o R.H.S. o (47) are he same or all he proocols. Hence T s only aeced by he esmaon error σ k and ncreasng wh σ k. As σ max, as >σ max, sod >σ max, sod-s, we have T, qla < T, sod < T, as. oe ha TCP-SoD s anoher verson o TCP-LA ha uses SoD mehod o esmae he queue lengh nsead o SoD-TS. Addonally, o model FAST TCP usng he same sysem model we assume FAST TCP adops a large nal CWnd accordng o (4) hus FAST TCP skps TCP slow sar phase and eners TCP congeson avodance phase drecly over whch FAST algorhm s appled. E. ueueng Delay I s a ac ha under he same bandwdh varaons he esmaon error σ ncreases wh he ncrease n uplnk delay varaon λ. Subsue T, qla, T, sod and T, as no (39) and le () qla, () as and () sod denoe he queue lengh o TCP-LA, TCP-SoD and FAST TCP a me, we ound ha () sod and () as decreases wh he ncrease n uplnk delay varaons, whle () qla s no aeced by he uplnk delay varaons, hence we have he ollowng heorem, THEOREM 3: under he same bandwdh varaons () qla, () as and () sod sasy he ollowng relaon: () qla () sod () as, and he equaly s acheved when operang under he maxmum uplnk delay varaons. Proo: Accordng o Theorem 2, he mnmum T or a proocol o sasy he ull bandwdh ulzaon or all s le T = max +max ou + D mn ou D mn (48) Subsue T n (46) wh (48) we have n ou Dmn ou Dmn max ou D mn ou D mn max When σ =σ max, we have n ou Dmn ou Dmn max ou D mn ou D mn Whch mples ha he queueng lengh s he same or every proocol he esmaon error reaches he maxma, e.g., σ =σ max. Snce he σ s ncreasng wh he uplnk delay varaon, we have () as =() sod =() qla when operang under he maxmum uplnk delay varaons. In he oher cases, e.g., σ <σ max we have () as >() sod >() qla as σ max, as >σ max, sod >σ max, sod-s. 567

7 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) recevng hroughpu (Mbps) 6 5 cp-qla = cp-sod =7 as cp =8.3 cp-qla =2 cp vegas =8.29 cp-sod =6 as cp =6 cp vegas =4 cp vegas = RTT (s) cp-sod =3 as cp =3 cp-qla =5.26 as cp =4 cp-sod =5 cp vegas =5 as cp = Fg. 2 Throughpu and RTT perormance o ve congeson conrol algorhms versus deren arge queue lengh. recevng hroughpu (Mbps) LA-3 hroughpu FAST-8 hroughpu Vegas-8 hroughpu Cubc hroughpu SoD-7 hroughpu LA-3 RTT FAST-8 RTT Vegas-8 RTT Cubc RTT SoD-7 RTT uplnk delay varaons (ms) Fg. 3 Throughpu and RTT perormance o ve congeson conrol algorhms versus deren levels o uplnk delay varaons (zero random loss). recevng hroughpu (Mbps) SoD-7 RTT Fas-8 hroughpu LA-3 RTT Cubc RTT LA-3 hroughpu SoD-7 hroughpu Fas-8 RTT.5.2 Vegas-8 RTT Cubc hroughpu Vegas-8 hroughpu uplnk delay varaon (ms) Fg. 4 Throughpu and RTT perormance o ve congeson conrol algorhms versus deren levels o uplnk delay varaons (.5% random loss). Theorem 3 shows ha TCP-LA always resuls n he lowes queueng delay under varous nework condons, whle sll ully ulze he bandwdh, whch s benecal or some applcaons requrng boh hgh hroughpu and low end-o-end delay such as vdeo conerence RTT (s) RTT (s) V. PERFORMAE EVALUATIOS In hs secon we evaluae he perormance o he proposed TCP-LA, and compare o oher TCP varans usng race-drven smulaons usng S2 [5] n he conex o modern 3G/HSPA nework. We mplemened he TCP-LA algorhm descrbed n Secon III n he S2 smulaor. We se up he nework opology n Fg. wh a lnk buer sze o 852 packes and a base propagaon delay o 6ms measured rom a producon o 3G/HSPA nework. To capure he acual bandwdh varaons n producon 3G/HSPA neworks, we measured he raw bandwdh o producon moble neworks usng UDP by sendng UDP daagrams rom a wred server o a 3G-conneced recever a a rae hgher han he moble lnk capacy. The receved daa rae race a he recever sde hus capures he acual bandwdh avalable over he rado lnk. These bandwdh races are hen ed no he smulaor o replcae he bandwdh varaons n he smulaed rado downlnk. For he uplnk we nroduce random delay varaons whch are unorm-dsrbued wh sandard devaon rangng rom o ms, approxmang he uplndk delay varaons observed n producon neworks. Each se o smulaon lased seconds and he resuls are averaged over runs wh deren bandwdh races. Frsly, we valdae he heorems derved rom he sysem model n Secon IV. Speccally, we plo he recevng hroughpu and RTT resuls o TCP varans usng deren T. To derve he mnmum requred T o every TCP varans or curren producon o 3G/HSPA nework, we eed he same bandwdh races no he smulaor or all cases. The average bandwdh o he races s 4.76Mbps. Addonally, we se he sandard devaon o uplnk delay varaons o he maxma, e.g., ms. As shown n Fg. 2, TCP-LA can acheve near % bandwdh ulzaon (e.g., 4.63/4.76=97.3%) wh T =3 packes, whle TCP-SoD and FAST TCP requres T =7 and 8 packes o acheve he smlar bandwdh ulzaon respecvely as hey have he larger queue lengh esmaon errors. Ths valdaes he Theorem 2 n Secon III, hus T, qla =3,, T, sod =7 and T, as =8. Generally, TCP-LA requres a smaller T han TCP-SoD and FAST TCP o acheve he smlar bandwdh ulzaon, e.g., TCP-LA acheves 7.2% wh T =5 packes, whle TCP-SoD and FAST TCP requre T =3 and 4 packes o acheve respecvely. On he oher hand, TCP-LA, TCP-SoD and FAST TCP resul n he smlar end-o-end delays, e.g., 3ms, usng T, qla,, T, sod and T, as respecvely, whch s shown by Theorem 3 ha hey wll resul n he same () under he maxmum uplnk delay varaons. We second compare he long-erm hroughpu and he RTT o ve TCP varans: (a) TCP-LA wh T =3 (TCP-LA-3); (b) TCP-SoD wh T =7 (TCP-SoD-7); (c) FAST TCP wh T =8 (FAST-TCP-7); (d) TCP Vegas wh T =8 (TCP-Vegas-8); and (e) TCP CUBIC. oe ha TCP varans n (a)-(d) adop he mnmum arge queue lengh derved rom Fg. 2. We vary he degree o uplnk delay varaons wh sandard devaon rangng rom o ms. As shown n Fg. 3, TCP-LA-3, TCP-SoD-7, FAST-TCP-8 568

8 24 IEEE h Inernaonal Conerence on Wreless and Moble Compung, eworkng and Communcaons (WMob) and TCP CUBIC all acheve near-% bandwdh ulzaon (e.g., over 4.63/4.76=97.2%) or all levels o uplnk delay varaons. However, TCP CUBIC resuls n he larges RTT close o.s. Ths s because TCP CUBIC keeps ncreasng he number o nlgh packes unl congeson loss s deeced, hus buldng up a large backlog n he lnk buer. Compared o TCP CUBIC, FAST-TCP-8, TCP-SoD-7 and TCP-LA-3 reduce he RTT a leas by a acor o 2.2 (.33s versus.6s) as hey all lm he queue lengh n he lnk buer. Among hese, TCP-LA-3 resuls n he smalles RTT or all levels o uplnk delay varaons as shown n Theorem 3, reducng RTT a mos by 5% compared o FAST-TCP-8 and TCP-SoD-7. TCP-Vegas-8 canno ecenly ulze he bandwdh snce ncreases/decrease CWnd by one every RTT, whch s shown o be conservave n ulzng he bandwdh and responsve slowly n conrollng he queue lengh. So ar he smulaons were perormed wh no random packe loss. We repeaed he smulaons wh a random packe loss rae o.5% and ploed he resuls n Fg. 4. We observe ha he hroughpu o TCP CUBIC and TCP Vegas degraded sgncanly n he presence o random packe losses. Ths s because hey boh decrease CWnd and sshresh upon loss evens, even non-congeson-nduced ones, hus unnecessarly reduced he ransmsson rae. By conras, FAST TCP does no decrease CWnd bu keep he CWnd consan durng he loss recovery hus FAST TCP has a beer hroughpu perormance han TCP CUBIC and TCP Vegas. Unlke hem, TCP LA and TCP SoD sll regulae he queue lengh around T usng (2) durng loss recovery (c.. Secon III.C) so ha CWnd s no aeced by he loss evens. As a resul hey manan smlar hroughpu and RTT perormances as he zero-random-loss cases despe he presence o random packe losses. Ths s sgncan as random packe losses are sll common n moble daa neworks. REFERECES [] M. Allman, V. Paxson, and W. Sevens, TCP Congeson Conrol, RFC 258, Aprl 999. [2] L. Brakmo and L. Peerson, TCP Vegas: End-o-end congeson avodance on a global nerne, IEEE JSAC, Ocober, 995. [3] D. X. We, C. Jn, S. H. Low, and S. Hegde, FAST TCP: movaon, archecure, algorhms, perormance, n IEEE/ACM TO, 26. [4] C. P. Fu and S. C. Lew, TCP Veno: TCP Enhancemen or Transmsson over Wreless Access eworks, n IEEE JSAC, 23. [5] Injong Rhee and Lsong Xu, CUBIC: A ew TCP-Frendly Hgh-Speed TCP Varans, PFLDne, 25 [6] E. Dahman, S. Parkvall, J. Skold, P. Bemng, 3G Evoluon: HSPA and LTE or Moble Broadband, 2nd ed., Academc Press, 28. [7] Haqng Jang, Yaogong Wang, yunghan Lee, and Injong Rhee, Tacklng buerbloa n 3G/4G neworks, n ACM IMC, 22. [8] Sanley C. F. Chan, e Lu,. M. Chan and Jack Y. B. Lee, On ueue Lengh and Lnk Buer Sze Esmaon n 3G/4G Moble Daa eworks, n IEEE TMC, 23. [9] Yn Xu, Wa ay Leong, Ben Leong and Al Razeen. "Dynamc Regulaon o Moble 3G/HSPA Uplnk Buer wh Recever-Sde Flow Conrol". n IEEE ICP, 22. [] e Lu and Jack Y. B. Lee, "Uplnk Delay Varaon Compensaon n ueue Lengh Esmaon over Moble Daa eworks", Acceped n ICCC, 24. [] V. Jacobson, R. Braden, and D. Borman, TCP exensons or hgh perormance, RFC 323, May 992. [2] R. Chakravory, S. a, I. Pra and J. Crowcro, Usng TCP Flow-Aggregaon o Enhance Daa Experence o Cellular Wreless Users, IEEE JSAC, Vol. 23, o. 6, June 25. [3] L. Bodrog, G. Horvah, and C. Vulkan, Analycal TCP hroughpu model or hgh-speed downlnk packe access, Soware, IET, 29. [4] S. Floyd, J. Mahdav, M. Mahs, and M. Podolsky, An exenson o he selecve acknowledgemen (SAC) opon or TCP, RFC 2883, 2.. [5] S2, ework Smulaor, hp:// VI. SUMMARY AD FUTURE WOR Ths work oers a novel queue-lengh-aware TCP (TCP-LA) ha regulaes he queue lengh a he boleneck lnk o a arge queue lengh by makng use o SoD-TS mehod. By analyss and smulaon we demonsrae ha TCP-LA resuls n he lowes end-o-end delay compared o loss-based TCP such as TCP CUBIC and delay-based TCP such as FAST TCP and TCP Vegas, whle sll acheves he near % bandwdh ulzaon. We are currenly exendng hs work n wo drecons. Frs, we are exendng he analycal model n Secon III o characerze TCP-LA and he oher delay-based TCP s overall bandwdh ecency. Second, we are mplemenng he TCP-LA algorhm no he Lnux kernel o evaluae s perormance over producons o moble daa neworks. VII. ACOWLEDGEMET Ths work was parally suppored by he Sraegc Prory Research Program o he Chnese Academy o Scences under Gran o. XDA64, by aonal aural Scence Foundaon o Chna (SFC) under Gran o , Gran o. 6338, and Gran o , and by Huawe Research Program YBCB

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS R&RATA # Vol.) 8, March FURTHER AALYSIS OF COFIDECE ITERVALS FOR LARGE CLIET/SERVER COMPUTER ETWORKS Vyacheslav Abramov School of Mahemacal Scences, Monash Unversy, Buldng 8, Level 4, Clayon Campus, Wellngon

More information

Relative controllability of nonlinear systems with delays in control

Relative controllability of nonlinear systems with delays in control Relave conrollably o nonlnear sysems wh delays n conrol Jerzy Klamka Insue o Conrol Engneerng, Slesan Techncal Unversy, 44- Glwce, Poland. phone/ax : 48 32 37227, {jklamka}@a.polsl.glwce.pl Keywor: Conrollably.

More information

CS 268: Packet Scheduling

CS 268: Packet Scheduling Pace Schedulng Decde when and wha pace o send on oupu ln - Usually mplemened a oupu nerface CS 68: Pace Schedulng flow Ion Soca March 9, 004 Classfer flow flow n Buffer managemen Scheduler soca@cs.bereley.edu

More information

12d Model. Civil and Surveying Software. Drainage Analysis Module Detention/Retention Basins. Owen Thornton BE (Mech), 12d Model Programmer

12d Model. Civil and Surveying Software. Drainage Analysis Module Detention/Retention Basins. Owen Thornton BE (Mech), 12d Model Programmer d Model Cvl and Surveyng Soware Dranage Analyss Module Deenon/Reenon Basns Owen Thornon BE (Mech), d Model Programmer owen.hornon@d.com 4 January 007 Revsed: 04 Aprl 007 9 February 008 (8Cp) Ths documen

More information

DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL

DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL Sco Wsdom, John Hershey 2, Jonahan Le Roux 2, and Shnj Waanabe 2 Deparmen o Elecrcal Engneerng, Unversy o Washngon, Seale, WA, USA

More information

Bandlimited channel. Intersymbol interference (ISI) This non-ideal communication channel is also called dispersive channel

Bandlimited channel. Intersymbol interference (ISI) This non-ideal communication channel is also called dispersive channel Inersymol nererence ISI ISI s a sgnal-dependen orm o nererence ha arses ecause o devaons n he requency response o a channel rom he deal channel. Example: Bandlmed channel Tme Doman Bandlmed channel Frequency

More information

Including the ordinary differential of distance with time as velocity makes a system of ordinary differential equations.

Including the ordinary differential of distance with time as velocity makes a system of ordinary differential equations. Soluons o Ordnary Derenal Equaons An ordnary derenal equaon has only one ndependen varable. A sysem o ordnary derenal equaons consss o several derenal equaons each wh he same ndependen varable. An eample

More information

Variants of Pegasos. December 11, 2009

Variants of Pegasos. December 11, 2009 Inroducon Varans of Pegasos SooWoong Ryu bshboy@sanford.edu December, 009 Youngsoo Cho yc344@sanford.edu Developng a new SVM algorhm s ongong research opc. Among many exng SVM algorhms, we wll focus on

More information

A Transparent Rate Adaptation Algorithm for Streaming Video over the Internet

A Transparent Rate Adaptation Algorithm for Streaming Video over the Internet A Transparen Rae Adapaon Algorhm for Sreamng Vdeo over he Inerne L. S. Lam, Jack Y. B. Lee, S. C. Lew, and W. Wang Deparmen of Informaon Engneerng The Chnese Unversy of Hong Kong Shan, N.T., Hong Kong,

More information

e-journal Reliability: Theory& Applications No 2 (Vol.2) Vyacheslav Abramov

e-journal Reliability: Theory& Applications No 2 (Vol.2) Vyacheslav Abramov June 7 e-ournal Relably: Theory& Applcaons No (Vol. CONFIDENCE INTERVALS ASSOCIATED WITH PERFORMANCE ANALYSIS OF SYMMETRIC LARGE CLOSED CLIENT/SERVER COMPUTER NETWORKS Absrac Vyacheslav Abramov School

More information

Increasing the Probablility of Timely and Correct Message Delivery in Road Side Unit Based Vehicular Communcation

Increasing the Probablility of Timely and Correct Message Delivery in Road Side Unit Based Vehicular Communcation Halmsad Unversy For he Developmen of Organsaons Producs and Qualy of Lfe. Increasng he Probablly of Tmely and Correc Message Delvery n Road Sde Un Based Vehcular Communcaon Magnus Jonsson Krsna Kuner and

More information

Time-interval analysis of β decay. V. Horvat and J. C. Hardy

Time-interval analysis of β decay. V. Horvat and J. C. Hardy Tme-nerval analyss of β decay V. Horva and J. C. Hardy Work on he even analyss of β decay [1] connued and resuled n he developmen of a novel mehod of bea-decay me-nerval analyss ha produces hghly accurae

More information

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany Herarchcal Markov Normal Mxure models wh Applcaons o Fnancal Asse Reurns Appendx: Proofs of Theorems and Condonal Poseror Dsrbuons John Geweke a and Gann Amsano b a Deparmens of Economcs and Sascs, Unversy

More information

Achieving High Throughput and Low Delay in Mobile Data Networks by Accurately Predicting Queue Lengths

Achieving High Throughput and Low Delay in Mobile Data Networks by Accurately Predicting Queue Lengths Achieving High Throughpu and Lo Delay in Mobile Daa eorks by Accuraely Predicing ueue Lenghs e Liu Sae ey Laboraory o Compuer Archiecure Insiue o Compuing Technology Chinese Academy o Sciences Email: liuke@ic.ac.cn

More information

( ) () we define the interaction representation by the unitary transformation () = ()

( ) () we define the interaction representation by the unitary transformation () = () Hgher Order Perurbaon Theory Mchael Fowler 3/7/6 The neracon Represenaon Recall ha n he frs par of hs course sequence, we dscussed he chrödnger and Hesenberg represenaons of quanum mechancs here n he chrödnger

More information

A Deterministic Algorithm for Summarizing Asynchronous Streams over a Sliding Window

A Deterministic Algorithm for Summarizing Asynchronous Streams over a Sliding Window A Deermnsc Algorhm for Summarzng Asynchronous Sreams over a Sldng ndow Cosas Busch Rensselaer Polyechnc Insue Srkana Trhapura Iowa Sae Unversy Oulne of Talk Inroducon Algorhm Analyss Tme C Daa sream: 3

More information

Outline. Probabilistic Model Learning. Probabilistic Model Learning. Probabilistic Model for Time-series Data: Hidden Markov Model

Outline. Probabilistic Model Learning. Probabilistic Model Learning. Probabilistic Model for Time-series Data: Hidden Markov Model Probablsc Model for Tme-seres Daa: Hdden Markov Model Hrosh Mamsuka Bonformacs Cener Kyoo Unversy Oulne Three Problems for probablsc models n machne learnng. Compung lkelhood 2. Learnng 3. Parsng (predcon

More information

[ ] 2. [ ]3 + (Δx i + Δx i 1 ) / 2. Δx i-1 Δx i Δx i+1. TPG4160 Reservoir Simulation 2018 Lecture note 3. page 1 of 5

[ ] 2. [ ]3 + (Δx i + Δx i 1 ) / 2. Δx i-1 Δx i Δx i+1. TPG4160 Reservoir Simulation 2018 Lecture note 3. page 1 of 5 TPG460 Reservor Smulaon 08 page of 5 DISCRETIZATIO OF THE FOW EQUATIOS As we already have seen, fne dfference appromaons of he paral dervaves appearng n he flow equaons may be obaned from Taylor seres

More information

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!) i+1,q - [(! ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL The frs hng o es n wo-way ANOVA: Is here neracon? "No neracon" means: The man effecs model would f. Ths n urn means: In he neracon plo (wh A on he horzonal

More information

Cubic Bezier Homotopy Function for Solving Exponential Equations

Cubic Bezier Homotopy Function for Solving Exponential Equations Penerb Journal of Advanced Research n Compung and Applcaons ISSN (onlne: 46-97 Vol. 4, No.. Pages -8, 6 omoopy Funcon for Solvng Eponenal Equaons S. S. Raml *,,. Mohamad Nor,a, N. S. Saharzan,b and M.

More information

Performance Analysis for a Network having Standby Redundant Unit with Waiting in Repair

Performance Analysis for a Network having Standby Redundant Unit with Waiting in Repair TECHNI Inernaonal Journal of Compung Scence Communcaon Technologes VOL.5 NO. July 22 (ISSN 974-3375 erformance nalyss for a Nework havng Sby edundan Un wh ang n epar Jendra Sngh 2 abns orwal 2 Deparmen

More information

Comparison of Differences between Power Means 1

Comparison of Differences between Power Means 1 In. Journal of Mah. Analyss, Vol. 7, 203, no., 5-55 Comparson of Dfferences beween Power Means Chang-An Tan, Guanghua Sh and Fe Zuo College of Mahemacs and Informaon Scence Henan Normal Unversy, 453007,

More information

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim Korean J. Mah. 19 (2011), No. 3, pp. 263 272 GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS Youngwoo Ahn and Kae Km Absrac. In he paper [1], an explc correspondence beween ceran

More information

Lecture 2 M/G/1 queues. M/G/1-queue

Lecture 2 M/G/1 queues. M/G/1-queue Lecure M/G/ queues M/G/-queue Posson arrval process Arbrary servce me dsrbuon Sngle server To deermne he sae of he sysem a me, we mus now The number of cusomers n he sysems N() Tme ha he cusomer currenly

More information

First-order piecewise-linear dynamic circuits

First-order piecewise-linear dynamic circuits Frs-order pecewse-lnear dynamc crcus. Fndng he soluon We wll sudy rs-order dynamc crcus composed o a nonlnear resse one-por, ermnaed eher by a lnear capacor or a lnear nducor (see Fg.. Nonlnear resse one-por

More information

Comb Filters. Comb Filters

Comb Filters. Comb Filters The smple flers dscussed so far are characered eher by a sngle passband and/or a sngle sopband There are applcaons where flers wh mulple passbands and sopbands are requred Thecomb fler s an example of

More information

Solution in semi infinite diffusion couples (error function analysis)

Solution in semi infinite diffusion couples (error function analysis) Soluon n sem nfne dffuson couples (error funcon analyss) Le us consder now he sem nfne dffuson couple of wo blocks wh concenraon of and I means ha, n a A- bnary sysem, s bondng beween wo blocks made of

More information

Department of Economics University of Toronto

Department of Economics University of Toronto Deparmen of Economcs Unversy of Torono ECO408F M.A. Economercs Lecure Noes on Heeroskedascy Heeroskedascy o Ths lecure nvolves lookng a modfcaons we need o make o deal wh he regresson model when some of

More information

Robustness Experiments with Two Variance Components

Robustness Experiments with Two Variance Components Naonal Insue of Sandards and Technology (NIST) Informaon Technology Laboraory (ITL) Sascal Engneerng Dvson (SED) Robusness Expermens wh Two Varance Componens by Ana Ivelsse Avlés avles@ns.gov Conference

More information

EG Low Voltage CMOS Fully Differential Current Feedback Amplifier with Controllable 3-dB Bandwidth

EG Low Voltage CMOS Fully Differential Current Feedback Amplifier with Controllable 3-dB Bandwidth EG0800330 Low olage CMS Fully Derenal Curren Feedback Ampler wh Conrollable 3dB Bandwdh Ahmed H. Madan 2, Mahmoud A. Ashour, Solman A. Mahmoud 2, and Ahmed M. Solman 3 adaon Engneerng Dep., NCT, EAEA Caro,

More information

Online Appendix for. Strategic safety stocks in supply chains with evolving forecasts

Online Appendix for. Strategic safety stocks in supply chains with evolving forecasts Onlne Appendx for Sraegc safey socs n supply chans wh evolvng forecass Tor Schoenmeyr Sephen C. Graves Opsolar, Inc. 332 Hunwood Avenue Hayward, CA 94544 A. P. Sloan School of Managemen Massachuses Insue

More information

HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD

HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD Journal of Appled Mahemacs and Compuaonal Mechancs 3, (), 45-5 HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD Sansław Kukla, Urszula Sedlecka Insue of Mahemacs,

More information

Efficient Asynchronous Channel Hopping Design for Cognitive Radio Networks

Efficient Asynchronous Channel Hopping Design for Cognitive Radio Networks Effcen Asynchronous Channel Hoppng Desgn for Cognve Rado Neworks Chh-Mn Chao, Chen-Yu Hsu, and Yun-ng Lng Absrac In a cognve rado nework (CRN), a necessary condon for nodes o communcae wh each oher s ha

More information

Inter-Class Resource Sharing using Statistical Service Envelopes

Inter-Class Resource Sharing using Statistical Service Envelopes In Proceedngs of IEEE INFOCOM 99 Iner-Class Resource Sharng usng Sascal Servce Envelopes Jng-yu Qu and Edward W. Knghly Deparmen of Elecrcal and Compuer Engneerng Rce Unversy Absrac Neworks ha suppor mulple

More information

CS286.2 Lecture 14: Quantum de Finetti Theorems II

CS286.2 Lecture 14: Quantum de Finetti Theorems II CS286.2 Lecure 14: Quanum de Fne Theorems II Scrbe: Mara Okounkova 1 Saemen of he heorem Recall he las saemen of he quanum de Fne heorem from he prevous lecure. Theorem 1 Quanum de Fne). Le ρ Dens C 2

More information

ON THE WEAK LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS

ON THE WEAK LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS ON THE WEA LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS FENGBO HANG Absrac. We denfy all he weak sequenal lms of smooh maps n W (M N). In parcular, hs mples a necessary su cen opologcal

More information

FTCS Solution to the Heat Equation

FTCS Solution to the Heat Equation FTCS Soluon o he Hea Equaon ME 448/548 Noes Gerald Reckenwald Porland Sae Unversy Deparmen of Mechancal Engneerng gerry@pdxedu ME 448/548: FTCS Soluon o he Hea Equaon Overvew Use he forward fne d erence

More information

On One Analytic Method of. Constructing Program Controls

On One Analytic Method of. Constructing Program Controls Appled Mahemacal Scences, Vol. 9, 05, no. 8, 409-407 HIKARI Ld, www.m-hkar.com hp://dx.do.org/0.988/ams.05.54349 On One Analyc Mehod of Consrucng Program Conrols A. N. Kvko, S. V. Chsyakov and Yu. E. Balyna

More information

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data Anne Chao Ncholas J Goell C seh lzabeh L ander K Ma Rober K Colwell and Aaron M llson 03 Rarefacon and erapolaon wh ll numbers: a framewor for samplng and esmaon n speces dversy sudes cology Monographs

More information

Linear Response Theory: The connection between QFT and experiments

Linear Response Theory: The connection between QFT and experiments Phys540.nb 39 3 Lnear Response Theory: The connecon beween QFT and expermens 3.1. Basc conceps and deas Q: ow do we measure he conducvy of a meal? A: we frs nroduce a weak elecrc feld E, and hen measure

More information

Dynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005

Dynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005 Dynamc Team Decson Theory EECS 558 Proec Shruvandana Sharma and Davd Shuman December 0, 005 Oulne Inroducon o Team Decson Theory Decomposon of he Dynamc Team Decson Problem Equvalence of Sac and Dynamc

More information

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4 CS434a/54a: Paern Recognon Prof. Olga Veksler Lecure 4 Oulne Normal Random Varable Properes Dscrmnan funcons Why Normal Random Varables? Analycally racable Works well when observaon comes form a corruped

More information

Mechanics Physics 151

Mechanics Physics 151 Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm H ( q, p, ) = q p L( q, q, ) H p = q H q = p H = L Equvalen o Lagrangan formalsm Smpler, bu

More information

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model BGC1: Survval and even hsory analyss Oslo, March-May 212 Monday May 7h and Tuesday May 8h The addve regresson model Ørnulf Borgan Deparmen of Mahemacs Unversy of Oslo Oulne of program: Recapulaon Counng

More information

Mechanics Physics 151

Mechanics Physics 151 Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm Hqp (,,) = qp Lqq (,,) H p = q H q = p H L = Equvalen o Lagrangan formalsm Smpler, bu wce as

More information

Genetic Algorithm in Parameter Estimation of Nonlinear Dynamic Systems

Genetic Algorithm in Parameter Estimation of Nonlinear Dynamic Systems Genec Algorhm n Parameer Esmaon of Nonlnear Dynamc Sysems E. Paeraks manos@egnaa.ee.auh.gr V. Perds perds@vergna.eng.auh.gr Ah. ehagas kehagas@egnaa.ee.auh.gr hp://skron.conrol.ee.auh.gr/kehagas/ndex.hm

More information

Chapters 2 Kinematics. Position, Distance, Displacement

Chapters 2 Kinematics. Position, Distance, Displacement Chapers Knemacs Poson, Dsance, Dsplacemen Mechancs: Knemacs and Dynamcs. Knemacs deals wh moon, bu s no concerned wh he cause o moon. Dynamcs deals wh he relaonshp beween orce and moon. The word dsplacemen

More information

Approximate Analytic Solution of (2+1) - Dimensional Zakharov-Kuznetsov(Zk) Equations Using Homotopy

Approximate Analytic Solution of (2+1) - Dimensional Zakharov-Kuznetsov(Zk) Equations Using Homotopy Arcle Inernaonal Journal of Modern Mahemacal Scences, 4, (): - Inernaonal Journal of Modern Mahemacal Scences Journal homepage: www.modernscenfcpress.com/journals/jmms.aspx ISSN: 66-86X Florda, USA Approxmae

More information

EP2200 Queuing theory and teletraffic systems. 3rd lecture Markov chains Birth-death process - Poisson process. Viktoria Fodor KTH EES

EP2200 Queuing theory and teletraffic systems. 3rd lecture Markov chains Birth-death process - Poisson process. Viktoria Fodor KTH EES EP Queung heory and eleraffc sysems 3rd lecure Marov chans Brh-deah rocess - Posson rocess Vora Fodor KTH EES Oulne for oday Marov rocesses Connuous-me Marov-chans Grah and marx reresenaon Transen and

More information

Graduate Macroeconomics 2 Problem set 5. - Solutions

Graduate Macroeconomics 2 Problem set 5. - Solutions Graduae Macroeconomcs 2 Problem se. - Soluons Queson 1 To answer hs queson we need he frms frs order condons and he equaon ha deermnes he number of frms n equlbrum. The frms frs order condons are: F K

More information

PHYS 1443 Section 001 Lecture #4

PHYS 1443 Section 001 Lecture #4 PHYS 1443 Secon 001 Lecure #4 Monda, June 5, 006 Moon n Two Dmensons Moon under consan acceleraon Projecle Moon Mamum ranges and heghs Reerence Frames and relae moon Newon s Laws o Moon Force Newon s Law

More information

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study)

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study) Inernaonal Mahemacal Forum, Vol. 8, 3, no., 7 - HIKARI Ld, www.m-hkar.com hp://dx.do.org/.988/mf.3.3488 New M-Esmaor Objecve Funcon n Smulaneous Equaons Model (A Comparave Sudy) Ahmed H. Youssef Professor

More information

Chapter 6: AC Circuits

Chapter 6: AC Circuits Chaper 6: AC Crcus Chaper 6: Oulne Phasors and he AC Seady Sae AC Crcus A sable, lnear crcu operang n he seady sae wh snusodal excaon (.e., snusodal seady sae. Complee response forced response naural response.

More information

A Data Aggregation Scheduling Algorithm with Longlifetime and Low-latency in Wireless Sensor Networks

A Data Aggregation Scheduling Algorithm with Longlifetime and Low-latency in Wireless Sensor Networks A Daa Aggregaon Schedulng Algorhm wh Longlfeme and Low-laency n Wreless Sensor eworks Zhengyu Chen,2,3 *, Geng Yang,2, Le Chen,2, Jn Wang 4, Jeong-Uk Km 5 Key Laboraory of Broad band Wreless Communcaon

More information

WiH Wei He

WiH Wei He Sysem Idenfcaon of onlnear Sae-Space Space Baery odels WH We He wehe@calce.umd.edu Advsor: Dr. Chaochao Chen Deparmen of echancal Engneerng Unversy of aryland, College Par 1 Unversy of aryland Bacground

More information

Tight results for Next Fit and Worst Fit with resource augmentation

Tight results for Next Fit and Worst Fit with resource augmentation Tgh resuls for Nex F and Wors F wh resource augmenaon Joan Boyar Leah Epsen Asaf Levn Asrac I s well known ha he wo smple algorhms for he classc n packng prolem, NF and WF oh have an approxmaon rao of

More information

TSS = SST + SSE An orthogonal partition of the total SS

TSS = SST + SSE An orthogonal partition of the total SS ANOVA: Topc 4. Orhogonal conrass [ST&D p. 183] H 0 : µ 1 = µ =... = µ H 1 : The mean of a leas one reamen group s dfferen To es hs hypohess, a basc ANOVA allocaes he varaon among reamen means (SST) equally

More information

SOME NOISELESS CODING THEOREMS OF INACCURACY MEASURE OF ORDER α AND TYPE β

SOME NOISELESS CODING THEOREMS OF INACCURACY MEASURE OF ORDER α AND TYPE β SARAJEVO JOURNAL OF MATHEMATICS Vol.3 (15) (2007), 137 143 SOME NOISELESS CODING THEOREMS OF INACCURACY MEASURE OF ORDER α AND TYPE β M. A. K. BAIG AND RAYEES AHMAD DAR Absrac. In hs paper, we propose

More information

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation Global Journal of Pure and Appled Mahemacs. ISSN 973-768 Volume 4, Number 6 (8), pp. 89-87 Research Inda Publcaons hp://www.rpublcaon.com Exsence and Unqueness Resuls for Random Impulsve Inegro-Dfferenal

More information

Journal of Theoretical and Applied Information Technology.

Journal of Theoretical and Applied Information Technology. Journal of heorecal and Appled Informaon echnology 5-9 JAI. All rghs reserved. www.ja.org NEW APPROXIMAION FOR ANDOFF RAE AND NUMBER OF ANDOFF PROBABILIY IN CELLULAR SYSEMS UNDER GENERAL DISRIBUIONS OF

More information

Range-based Mobility Estimations in MANETs with Application to Link Availability Prediction

Range-based Mobility Estimations in MANETs with Application to Link Availability Prediction Range-based Mobly Esmaons n MANETs wh Applcaon o Ln Avalably Predcon Zhuoqun L, Lngfen Sun and Emmanuel C. Ifeachor School of Compung, Communcaons and Elecroncs Unversy of Plymouh Drae Crcus, Plymouh,

More information

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s Ordnary Dfferenal Equaons n Neuroscence wh Malab eamples. Am - Gan undersandng of how o se up and solve ODE s Am Undersand how o se up an solve a smple eample of he Hebb rule n D Our goal a end of class

More information

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION INTERNATIONAL TRADE T. J. KEHOE UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 27 EXAMINATION Please answer wo of he hree quesons. You can consul class noes, workng papers, and arcles whle you are workng on he

More information

This document is downloaded from DR-NTU, Nanyang Technological University Library, Singapore.

This document is downloaded from DR-NTU, Nanyang Technological University Library, Singapore. Ths documen s downloaded from DR-NTU, Nanyang Technologcal Unversy Lbrary, Sngapore. Tle A smplfed verb machng algorhm for word paron n vsual speech processng( Acceped verson ) Auhor(s) Foo, Say We; Yong,

More information

Networked Estimation with an Area-Triggered Transmission Method

Networked Estimation with an Area-Triggered Transmission Method Sensors 2008, 8, 897-909 sensors ISSN 1424-8220 2008 by MDPI www.mdp.org/sensors Full Paper Neworked Esmaon wh an Area-Trggered Transmsson Mehod Vnh Hao Nguyen and Young Soo Suh * Deparmen of Elecrcal

More information

Fall 2010 Graduate Course on Dynamic Learning

Fall 2010 Graduate Course on Dynamic Learning Fall 200 Graduae Course on Dynamc Learnng Chaper 4: Parcle Flers Sepember 27, 200 Byoung-Tak Zhang School of Compuer Scence and Engneerng & Cognve Scence and Bran Scence Programs Seoul aonal Unversy hp://b.snu.ac.kr/~bzhang/

More information

Studies of Emergency Evacuation Strategies based on Kinematic Wave Models of Network Vehicular Traffic

Studies of Emergency Evacuation Strategies based on Kinematic Wave Models of Network Vehicular Traffic Sudes o Emergency Evacuaon Sraeges based on Knemac Wave Models o Nework Vehcular Trac KAI-FU QIU Deparmen o Auomaon, Unversy o Scence and Technology o Chna Hee, Chna 2327 prc@mal.usc.edu.cn Absrac- How

More information

CHAPTER 10: LINEAR DISCRIMINATION

CHAPTER 10: LINEAR DISCRIMINATION CHAPER : LINEAR DISCRIMINAION Dscrmnan-based Classfcaon 3 In classfcaon h K classes (C,C,, C k ) We defned dscrmnan funcon g j (), j=,,,k hen gven an es eample, e chose (predced) s class label as C f g

More information

Implementation of Quantized State Systems in MATLAB/Simulink

Implementation of Quantized State Systems in MATLAB/Simulink SNE T ECHNICAL N OTE Implemenaon of Quanzed Sae Sysems n MATLAB/Smulnk Parck Grabher, Mahas Rößler 2*, Bernhard Henzl 3 Ins. of Analyss and Scenfc Compung, Venna Unversy of Technology, Wedner Haupsraße

More information

NATIONAL UNIVERSITY OF SINGAPORE PC5202 ADVANCED STATISTICAL MECHANICS. (Semester II: AY ) Time Allowed: 2 Hours

NATIONAL UNIVERSITY OF SINGAPORE PC5202 ADVANCED STATISTICAL MECHANICS. (Semester II: AY ) Time Allowed: 2 Hours NATONAL UNVERSTY OF SNGAPORE PC5 ADVANCED STATSTCAL MECHANCS (Semeser : AY 1-13) Tme Allowed: Hours NSTRUCTONS TO CANDDATES 1. Ths examnaon paper conans 5 quesons and comprses 4 prned pages.. Answer all

More information

Lecture 11 SVM cont

Lecture 11 SVM cont Lecure SVM con. 0 008 Wha we have done so far We have esalshed ha we wan o fnd a lnear decson oundary whose margn s he larges We know how o measure he margn of a lnear decson oundary Tha s: he mnmum geomerc

More information

Panel Data Regression Models

Panel Data Regression Models Panel Daa Regresson Models Wha s Panel Daa? () Mulple dmensoned Dmensons, e.g., cross-secon and me node-o-node (c) Pongsa Pornchawseskul, Faculy of Economcs, Chulalongkorn Unversy (c) Pongsa Pornchawseskul,

More information

Notes on the stability of dynamic systems and the use of Eigen Values.

Notes on the stability of dynamic systems and the use of Eigen Values. Noes on he sabl of dnamc ssems and he use of Egen Values. Source: Macro II course noes, Dr. Davd Bessler s Tme Seres course noes, zarads (999) Ineremporal Macroeconomcs chaper 4 & Techncal ppend, and Hamlon

More information

Advanced time-series analysis (University of Lund, Economic History Department)

Advanced time-series analysis (University of Lund, Economic History Department) Advanced me-seres analss (Unvers of Lund, Economc Hsor Dearmen) 3 Jan-3 Februar and 6-3 March Lecure 4 Economerc echnues for saonar seres : Unvarae sochasc models wh Box- Jenns mehodolog, smle forecasng

More information

Sampling Procedure of the Sum of two Binary Markov Process Realizations

Sampling Procedure of the Sum of two Binary Markov Process Realizations Samplng Procedure of he Sum of wo Bnary Markov Process Realzaons YURY GORITSKIY Dep. of Mahemacal Modelng of Moscow Power Insue (Techncal Unversy), Moscow, RUSSIA, E-mal: gorsky@yandex.ru VLADIMIR KAZAKOV

More information

Li An-Ping. Beijing , P.R.China

Li An-Ping. Beijing , P.R.China A New Type of Cpher: DICING_csb L An-Png Bejng 100085, P.R.Chna apl0001@sna.com Absrac: In hs paper, we wll propose a new ype of cpher named DICING_csb, whch s derved from our prevous sream cpher DICING.

More information

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue.

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue. Lnear Algebra Lecure # Noes We connue wh he dscusson of egenvalues, egenvecors, and dagonalzably of marces We wan o know, n parcular wha condons wll assure ha a marx can be dagonalzed and wha he obsrucons

More information

Computing Relevance, Similarity: The Vector Space Model

Computing Relevance, Similarity: The Vector Space Model Compung Relevance, Smlary: The Vecor Space Model Based on Larson and Hears s sldes a UC-Bereley hp://.sms.bereley.edu/courses/s0/f00/ aabase Managemen Sysems, R. Ramarshnan ocumen Vecors v ocumens are

More information

Math 128b Project. Jude Yuen

Math 128b Project. Jude Yuen Mah 8b Proec Jude Yuen . Inroducon Le { Z } be a sequence of observed ndependen vecor varables. If he elemens of Z have a on normal dsrbuon hen { Z } has a mean vecor Z and a varancecovarance marx z. Geomercally

More information

Life-Add: Lifetime Adjustable Design for WiFi Networks with Heterogeneous Energy Supplies

Life-Add: Lifetime Adjustable Design for WiFi Networks with Heterogeneous Energy Supplies Lfe-Add: Lfeme Adjusable Desgn for WF Neworks wh Heerogeneous Energy Supples Shengbo Chen, Tarun Bansal, Yn Sun, Prasun Snha and Ness B. Shroff Deparmen of ECE, The Oho Sae Unversy Deparmen of CSE, The

More information

Multi-priority Online Scheduling with Cancellations

Multi-priority Online Scheduling with Cancellations Submed o Operaons Research manuscrp (Please, provde he manuscrp number!) Auhors are encouraged o subm new papers o INFORMS journals by means of a syle fle emplae, whch ncludes he journal le. However, use

More information

. The geometric multiplicity is dim[ker( λi. A )], i.e. the number of linearly independent eigenvectors associated with this eigenvalue.

. The geometric multiplicity is dim[ker( λi. A )], i.e. the number of linearly independent eigenvectors associated with this eigenvalue. Mah E-b Lecure #0 Noes We connue wh he dscusson of egenvalues, egenvecors, and dagonalzably of marces We wan o know, n parcular wha condons wll assure ha a marx can be dagonalzed and wha he obsrucons are

More information

Example: MOSFET Amplifier Distortion

Example: MOSFET Amplifier Distortion 4/25/2011 Example MSFET Amplfer Dsoron 1/9 Example: MSFET Amplfer Dsoron Recall hs crcu from a prevous handou: ( ) = I ( ) D D d 15.0 V RD = 5K v ( ) = V v ( ) D o v( ) - K = 2 0.25 ma/v V = 2.0 V 40V.

More information

Using Fuzzy Pattern Recognition to Detect Unknown Malicious Executables Code

Using Fuzzy Pattern Recognition to Detect Unknown Malicious Executables Code Usng Fuzzy Paern Recognon o Deec Unknown Malcous Execuables Code Boyun Zhang,, Janpng Yn, and Jngbo Hao School of Compuer Scence, Naonal Unversy of Defense Technology, Changsha 40073, Chna hnxzby@yahoo.com.cn

More information

Robust and Accurate Cancer Classification with Gene Expression Profiling

Robust and Accurate Cancer Classification with Gene Expression Profiling Robus and Accurae Cancer Classfcaon wh Gene Expresson Proflng (Compuaonal ysems Bology, 2005) Auhor: Hafeng L, Keshu Zhang, ao Jang Oulne Background LDA (lnear dscrmnan analyss) and small sample sze problem

More information

A New Rate Control Algorithm for MPEG-4 Video Coding

A New Rate Control Algorithm for MPEG-4 Video Coding A New Rae onrol Algorhm for MPEG-4 Vdeo odng Sun Yu and Ishfaq Ahmad Deparmen of ompuer Scence Hong Kong Unversy of Scence and Technology cssunyu@cs.us.hk, ahmad@cs.us.hk ABSTRAT Ths paper proposes a new

More information

CH.3. COMPATIBILITY EQUATIONS. Continuum Mechanics Course (MMC) - ETSECCPB - UPC

CH.3. COMPATIBILITY EQUATIONS. Continuum Mechanics Course (MMC) - ETSECCPB - UPC CH.3. COMPATIBILITY EQUATIONS Connuum Mechancs Course (MMC) - ETSECCPB - UPC Overvew Compably Condons Compably Equaons of a Poenal Vecor Feld Compably Condons for Infnesmal Srans Inegraon of he Infnesmal

More information

Volatility Interpolation

Volatility Interpolation Volaly Inerpolaon Prelmnary Verson March 00 Jesper Andreasen and Bran Huge Danse Mares, Copenhagen wan.daddy@danseban.com brno@danseban.com Elecronc copy avalable a: hp://ssrn.com/absrac=69497 Inro Local

More information

Appendix to Online Clustering with Experts

Appendix to Online Clustering with Experts A Appendx o Onlne Cluserng wh Expers Furher dscusson of expermens. Here we furher dscuss expermenal resuls repored n he paper. Ineresngly, we observe ha OCE (and n parcular Learn- ) racks he bes exper

More information

Supplementary Material to: IMU Preintegration on Manifold for E cient Visual-Inertial Maximum-a-Posteriori Estimation

Supplementary Material to: IMU Preintegration on Manifold for E cient Visual-Inertial Maximum-a-Posteriori Estimation Supplemenary Maeral o: IMU Prenegraon on Manfold for E cen Vsual-Ineral Maxmum-a-Poseror Esmaon echncal Repor G-IRIM-CP&R-05-00 Chrsan Forser, Luca Carlone, Fran Dellaer, and Davde Scaramuzza May 0, 05

More information

Should Exact Index Numbers have Standard Errors? Theory and Application to Asian Growth

Should Exact Index Numbers have Standard Errors? Theory and Application to Asian Growth Should Exac Index umbers have Sandard Errors? Theory and Applcaon o Asan Growh Rober C. Feensra Marshall B. Rensdorf ovember 003 Proof of Proposon APPEDIX () Frs, we wll derve he convenonal Sao-Vara prce

More information

Pendulum Dynamics. = Ft tangential direction (2) radial direction (1)

Pendulum Dynamics. = Ft tangential direction (2) radial direction (1) Pendulum Dynams Consder a smple pendulum wh a massless arm of lengh L and a pon mass, m, a he end of he arm. Assumng ha he fron n he sysem s proporonal o he negave of he angenal veloy, Newon s seond law

More information

Lecture 6: Learning for Control (Generalised Linear Regression)

Lecture 6: Learning for Control (Generalised Linear Regression) Lecure 6: Learnng for Conrol (Generalsed Lnear Regresson) Conens: Lnear Mehods for Regresson Leas Squares, Gauss Markov heorem Recursve Leas Squares Lecure 6: RLSC - Prof. Sehu Vjayakumar Lnear Regresson

More information

On computing differential transform of nonlinear non-autonomous functions and its applications

On computing differential transform of nonlinear non-autonomous functions and its applications On compung dfferenal ransform of nonlnear non-auonomous funcons and s applcaons Essam. R. El-Zahar, and Abdelhalm Ebad Deparmen of Mahemacs, Faculy of Scences and Humanes, Prnce Saam Bn Abdulazz Unversy,

More information

WebAssign HW Due 11:59PM Tuesday Clicker Information

WebAssign HW Due 11:59PM Tuesday Clicker Information WebAssgn HW Due 11:59PM Tuesday Clcker Inormaon Remnder: 90% aemp, 10% correc answer Clcker answers wll be a end o class sldes (onlne). Some days we wll do a lo o quesons, and ew ohers Each day o clcker

More information

FI 3103 Quantum Physics

FI 3103 Quantum Physics /9/4 FI 33 Quanum Physcs Aleander A. Iskandar Physcs of Magnesm and Phooncs Research Grou Insu Teknolog Bandung Basc Conces n Quanum Physcs Probably and Eecaon Value Hesenberg Uncerany Prncle Wave Funcon

More information

Density Matrix Description of NMR BCMB/CHEM 8190

Density Matrix Description of NMR BCMB/CHEM 8190 Densy Marx Descrpon of NMR BCMBCHEM 89 Operaors n Marx Noaon If we say wh one bass se, properes vary only because of changes n he coeffcens weghng each bass se funcon x = h< Ix > - hs s how we calculae

More information

Density Matrix Description of NMR BCMB/CHEM 8190

Density Matrix Description of NMR BCMB/CHEM 8190 Densy Marx Descrpon of NMR BCMBCHEM 89 Operaors n Marx Noaon Alernae approach o second order specra: ask abou x magnezaon nsead of energes and ranson probables. If we say wh one bass se, properes vary

More information

Let s treat the problem of the response of a system to an applied external force. Again,

Let s treat the problem of the response of a system to an applied external force. Again, Page 33 QUANTUM LNEAR RESPONSE FUNCTON Le s rea he problem of he response of a sysem o an appled exernal force. Agan, H() H f () A H + V () Exernal agen acng on nernal varable Hamlonan for equlbrum sysem

More information