Dynamic Multi-Level Capacitated and Uncapacitated Location Problems: an approach using primal-dual heuristics

Size: px
Start display at page:

Download "Dynamic Multi-Level Capacitated and Uncapacitated Location Problems: an approach using primal-dual heuristics"

Transcription

1 Dynamc Mul-Leel Caacaed and Uncaacaed Locaon Problems: an aroach usng rmal-dual heurscs JOANA DIAS (), M. EUGÉNIA CAPIVO () AND JOÃO CLÍMACO () ()Faculdade de Economa and INESC-Combra Unersdade de Combra A. Das da Sla, Combra Porugal () Unersdade de Lsboa, Faculdade de Cêncas Cenro de Inesgação Oeraconal Camo Grande, Bloco C6, Pso Lsboa Porugal Absrac: In hs aer seeral dynamc mul-leel locaon roblems are formulaed as mxed-neger lnear rograms. Boh uncaacaed and caacaed ersons of he roblem are suded. he models resened are more comlee han he ones known from he leraure: hey are dynamc and consder he ossbly of a facly beng oen, closed and reoen more han once durng he lannng horzon. hey may nclude boh uer and lower lms on he used caacy of each facly and may also consder he suaon where here s no flow conseraon n he nermedae facles. Prmal-dual heurscs were deeloed o sole effcenly he roosed models. Comuaonal resuls are resened and dscussed. Inroducon Mul-leel locaon roblems hae been wdely suded n he leraure. Auhors sudyng and wrng abou hese roblems desgnae hem dfferenly: herarchcal, mul-leel, mul-echelon, mul-sage, ec. Generally, he desgnaon of he roblem ndcaes he maxmum number of leels consdered: k-herarchcal, k-leel, k-echelon locaon roblems refer o roblems wh, a mos, k leels of facles. he models deeloed n hs research reor are desgnaed by mul-leel, because s he auhors undersandng ha hs s he mos general desgnaon (a mul-leel locaon roblem may no necessarly reresen a herarchcal locaon roblem; n he laer case facles hae o be organzed n a herarchcal srucure, whle n he frs case hs s no necessarly he case). hs research was arally suored by research roec POCI/ISFL-/5 and POCI/MA/39/00.

2 here are seeral examles n our daly les ha show he morance of consderng mul-leel facly locaon roblems: he herarchcal healh serce sysem, he herarchcal educaon sysem, he mul-leel srucure of bank and os-offces organzaons, ec. hese facly srucures hae some common feaures and some moran dfferenang characerscs. One can say ha a mul-leel facly locaon model s needed wheneer he facles o be locaed can be groued n ses (leels) hang dfferen characerscs (by offerng dfferen serces o clens, for nsance) and nerac wh each oher so ha s no ossble o locae facles n each se ndeendenly from he ohers. he exsng lnks beween he dfferen knds of facles can be mlc (n he form of a consran ha lms he global budge ha can be sen globally), or exlcly deermnng he flow of clens beween facles (Daskn, 995). Narula (984) aems o classfy he herarchcal locaon-allocaon roblems. He dfferenaes he roblems hrough he facly herarchy (successely ncluse a facly n leel m offers all serces of facles n leels o m- lus serces of leel m; successely excluse a facly n leel m offers serces unque o ); arc and node flow dsclne (flow s negraed f s from any lower leel o any hgher leel or s dscrmnang f he flow s from leel m o leel m; a unah nework s a nework where he ose degree of eery node s less han or equal o one, whereas n a mulah nework he ose degree of a leas one node s greaer han one). o comleely secfy an herarchcal locaon-allocaon roblem s needed o sae he number of leels, he ye of facly herarchy, he arc and flow dsclne. Neerheless he classfcaon scheme resened s no exhause, as he auhor remarks. Daskn (995, ) ges he basc noons of herarchcal facles, and descrbes some models: medan-based, coerage-based herarchcal locaon formulaon and some exensons o hose cases. he models descrbed n he leraure are, almos exclusely, sac roblems. Some consder he uncaacaed ohers he caacaed erson of he mul-leel roblem. Moore and ReVelle (98) descrbe a real case of successely ncluse herarchcal serce locaon roblem. hey consder as obece he mnmzaon of he oulaon ha lacks access o one or more comonens of serce. A oulaon s coered by a gen leel of serce f some member of he facly herarchy elgble o rode ha serce s locaed whn a maxmum esablshed dsance. he lnear relaxaon of he roblem s soled, followed by a branch and bound rocedure wheneer necessary. en e al (983) deelo and sole wo models wh alcaons n he herarchcal healh facly locaon-allocaon roblem. cha and Lee (984) resen a branch and bound algorhm based on a dual ascen rocedure (smlar o he one deeloed by Erlenkoer, 978) for he mul-leel uncaacaed facly locaon roblem. en years laer, a aer by Barros and Labbé (994) roe ha he mul-leel uncaacaed facly locaon roblem s no submodular, and queson he aldy of he branch and bound rocedure descrbed n cha and Lee

3 (984). Ro and cha (984) descrbe a branch and bound algorhm for resoluon of he wo-leel uncaacaed facly locaon wh addonal consrans ha reresen he adunc relaonsh of some warehouses o a ceran lan (f a lan s oen, here s a se of warehouses ha hae o be also oen). Narula and Ogbu (985) use lagrangean relaxaon wh subgraden omsaon o sole a wo-herarchcal uncaacaed successely ncluse locaon roblem. he model consders he locaon of facles n leel and facles n leel. Only a fracon of he demand from leel facles s referred o leel facles. Gao and Robnson (99) nesgae he use of dual-based rocedures smlar o he ones deelo by Erlenkoer (978) n he resoluon of he wo-echelon uncaacaed facly locaon roblem. Aardal e al (996) nesgae srucural roeres of he uncaacaed wo-leel facly locaon roblem. wo dfferen formulaons are suded, ha use dfferen decson arables (he auhors call hese wo formulaons he sngle and he mul commody flow formulaons). Edwards (00) sudes seeral roeres and descrbes aroxmaon algorhms for he mul-leel facly locaon roblem. Bumb (00) and Zhang and Ye (00) descrbe aroxmaon algorhms for he wo leel uncaacaed facly locaon roblem wh a maxmzaon obece. Galão e al (00) reas a 3-leel successely ncluse facly locaon model aled o a case sudy. here s an uer bound on he maxmum number of facles o locae a each leel k, k=,,3. he auhors do no consder he exsence of fxed oenng coss. wo heurscs are deeloed o fnd feasble soluons o he roblem. Eseo e al (003) rea he maxmal coerng wo-leel locaon roblem. A combned Lagrangean-surrogae relaxaon s mbedded no a subgraden omzaon algorhm o fnd lower and uer bounds o he omal soluon. Some references on caacaed mul-leel roblems are also aalable. Ean e al (99) resen a mxed-neger lnear rogrammng model, whch allows dfferen herarchcal relaonshs o be smulaneously resen, caacy consrans o be laced boh on serce yes and serce grous and he consderaon of boh fxed and arable coss. he model s aled o seeral leraure-based roblems and also o a new large-scale roblem ha fully llusraes he model caables. Bloemhof-Ruwaard e al (996) resen a caacaed erson of he wo-leel facly locaon roblem. hey consder wo dfferen model formulaons, and comare he qualy of he lower bounds obaned by her lnear relaxaons wh he lower bounds obaned wh a lagrangean relaxaon. he auhors use LP round-off heurscs and also sequenal caacaed facly locaon heurscs o fnd feasble rmal soluons. raganalerngsak e al (997) deelo and comare sx dfferen lagrangean relaxaons o he wo-echelon, sngle source caacaed facly locaon roblem. he lagrangean relaxaons are used whn a subgraden omsaon algorhm, and feasble soluons are consruced by heurscs based on a general assgnmen roblem formulaon. In a followng aer (raganalerngsak e al, 000), a branch and bound 3

4 mehod based on he mos effcen lagrangean relaxaon s resened. Prkul and Jayaraman (998) deelo a mul-commody, caacaed, wo-leel facly locaon model. hey consder he locaon of lans and warehouses. he roblem s soled heurscally, consrucng a rmal feasble soluon from he soluon of a lagrangan relaxaon (whn a subgraden mehod omsaon). Chardare (999) reas boh uncaacaed and caacaed wo-leel locaon roblems n he elecommuncaons nework-lannng conex. Lagrangean relaxaon and smulaed annealng are boh used o fnd feasble soluons. Klose (995, 998, 000) consders a wo-leel sngle sourcng caacaed facly locaon roblem (clens are sered from deos ha, n urn, are sered by lans). he locaon decson arables consder only he roblem of locang deos ( s consdered ha lans are already locaed a fxed ses). he roblem s soled usng lagrangean relaxaon followed by a heursc rocedure. he auhor res o sole he same roblem usng a LP-based heursc (Klose, 999): nroducng seeral ald nequales moaed by he subroblems conaned n he mul-leel roblem, he LP relaxaon s srenghen and soled. A heursc rocedure bulds an admssble soluon from he omal lnear soluon calculaed. Seeral comuaonal ess where erformed o assess he erformance of boh resoluon aroaches. Jayaraman e al (003) consder a herarchcal facly srucure where clens hae wo dfferen knds of demand: demand for basc and secalzed serces. he model resened nends o maxmze he oal demand coerage, allocang leels of serces o oen facles and consderng caacy consrans. A lagrangean relaxaon followed by a rmal heursc s deeloed, and he ably of he aroach s demonsraed hrough he resuls of exense comuaonal exermens. Dynamc mul-leel locaon roblems are descrbed n a few number of more recen aers. Melachrnouds and Mn (000) resen a real case of dynamc relocang an exsng facly ha belongs o a wo-leel facles srucure. he ranson beween he exsen and he new locaon has o be done gradually. he auhors consder boh caacy and budge consrans and more han one obece funcon. No dedcaed rocedure s deeloed: he mul-obece mxed-neger lnear roblem s soled usng a general soler. Hnoosa e al (000) model he dynamc woechelon mulcommody caacaed lan locaon roblem. he model consders he ossbly of oenng a facly a he begnnng of any me erod (remanng oen unl he end of he lannng horzon), and closng an already exsng facly a he end of any me erod (remanng closed unl he end of he lannng horzon). he calculaon of admssble soluons s done usng heurscs based on a lagrangean relaxaon. Canel e al (00) deelo a model for he same knd of roblem, bu consder he ossbly of a facly beng oen n more han one me erod, no sequenally. he auhors consder reoenng and closng coss, bu resen a non-lnear obece funcon. he aer descrbes an algorhm o sole he roblem. 4

5 here are oher neresng asecs of mul-leel facly locaon roblems ha hae also been reaed by some auhors. Hodgson (98) sudes a dfferen roblem ha can be consdered a knd of mul-leel locaon roblem: he locaon of ublc facles nermedae o he ourney o work. I s consdered ha clens hae o rael daly from home o work. he obece s o locae ublc facles such ha he exra me needed o rael o hose facles s mnmzed. Madsen (983) sudes he roblem of combned locaon-roung roblems n a sysem comosed of clens, deos and, a mos, one facory. Serra e al (99) sudy he roblem of locang facles wh a herarchcal srucure when here s comeon n he regon of neres. he model deeloed allows for boh he locaon of new facles and he relocaon of exsng facles. Marano and Serra (00) sudy a herarchcal locaon-allocaon model where he congeson roblem s reaed exlcly: here are suaons where clens hae o wa on queue for some me before beng sered. he auhors descrbe a b-leel heursc for consrucon of feasble soluons, and resen he resuls of comuaonal ess erformed. In hs research reor seeral dynamc mul-leel locaon roblems are addressed. In secon he uncaacaed erson of he roblem s descrbed, n secon 3 he caacaed erson wh maxmum caacy resrcons and, n secon 4, he caacaed erson wh maxmum and mnmum caacy resrcons bu wh no flow conseraon a he nermedae facles. In secons o 4, rmal-dual heurscs are descrbed ha can calculae admssble soluons for he corresondng roblem, based on he work of Erlenkoer (978), Van Roy and Erlenkoer (98) and Gugnard and Selberg (979). All roblems reaed n hs research reor hae he followng characerscs ha dsngush hem from he roblems usually referred o n he leraure:. he roblems are dynamc and consder ha a facly can be oen, closed and reoen more han once durng he lannng horzon. I s ossble o consder exlcly dfferen fxed oenng and reoenng coss (ha are, mos of he mes, clearly dfferen). I s ossble o consder he exsence of already oen facles. I s ossble o esablsh mnmum funconng me nerals for a gen facly (n some suaons, a facly ha s oen n me erod should be oen for a mnmum number of me erods before beng closed).. he models can be used for boh successely ncluse and excluse herarchcal roblems. he model s also ald for all he arcs and nodes flow dsclnes defned n Narula (984). 3. he model can deal wh ahs of facles wh a dfferen number of arcs. In a k-leel facly locaon roblem, s ossble o consder ahs wh a number of arcs from o k. hs s an neresng feaure for boh successely ncluse and excluse roblems. Imagne, for nsance, a healh facly serce sysem. Paens are adsed o go o her neares healh cenre, from here hey may be sen o a regonal hosal, and from here o a cenral hosal (wheneer s 5

6 needed). Somemes aens refer o go drecly o regonal or cenral hosals. In he models here resened, all hese suaons can be consdered. 4. I s ossble o consder more han one ah consued by exacly he same facles, bu wh dfferen characerscs. hs s an moran feaure, esecally f dealng wh more han one obece funcon. Imagne a roblem where here are wo obeces: mnmzng oal cos and oal rael me. Consder a ah consued by locaons. One can consder he locaons conneced by a hghway and by a naonal road. he frs oon wll be more exense bu faser, he second oon wll be less exense bu slower. Boh ahs can be smulaneously consdered n he model formulaon. In secon 5 some conclusons and fuure work drecons are addressed. Mul-Leel Uncaacaed Locaon Problem Fgure llusraes a ossble -leel nework confguraon. Clens are assgned o ahs, and no o sngle facles Fgure : Examle of a -leel facly nework. he symbol reresens clens, and reresens facles. Consder he followng defnons: J = {,...,,, n} se of ndexes corresondng o he clens locaons; I = {,...,,, m} se of ndexes corresondng o facles ossble locaons; P = {,, q } se of all ossble ahs; P( ) = { P: belongs o ah }; = number of me erods consdered n he lannng horzon ( ); K = maxmum number of leels n he mul-leel facly srucure; K wll also be he maxmum ah lengh. 6

7 c = cos of fully assgnng clen o ah n erod ; FA = fxed cos of oenng a facly a he begnnng of erod, and closng a he end of erod (he facly wll be n oeraon from he begnnng of o he end of ); FR = fxed cos of reoenng a facly a he begnnng of erod, and closng a he end of erod (he facly wll be n oeraon from he begnnng of o he end of ); and le us defne he arables: a r f facly s oen a he begnnng of erod and says oen unl he end of erod = 0 oherwse f facly s reoen a he begnnng of erod and says oen unl he end of erod =, > 0 oherwse f clen s assgned o ah durng erod x =. 0 oherwse he defnon of arables x was moaed by he work of cha and Lee (984). he man dfference has o do wh he fac ha each ah can hae any number of arcs from o K, whle n he referred o aer all ahs are of exacly he same lengh. Each ah can be consued by one or more facles. Each facly can belong o one or more ahs. Consder he wo-leel examle deced n fgure. All ahs reresened are ald. A ah wll be reresened by an ordered se (,,, k ), wh k K. If clen s assgned o ah, hs means ha clen s sered by facly, hen goes o facly, and so on, unl reachng facly k. Defnon : A ah s sad o be oen durng erod f and only f all facles belongng o are oen durng erod. Defnon : A ah s sad o be arally oen f and only f s no oen and a leas one facly belongng o s oen durng erod. Defnon 3: A ah s sad o be closed durng erod f and only f all facles belongng o are closed durng erod. I s no necessary o consder all ossble alues for (,),. If, for nsance, a facly has o be oerang for, a leas, s me erods erods afer beng oen, hen only ars (,) wh s should be consdered. 7

8 . Prmal Problem he dynamc, uncaacaed, mul-leel locaon roblem can be formulaed as DUMLP: DUMLP Mn c x = FA a > = FR r ( ) subec o: x =,, ( ) ( a r ) x 0 = = = = a = =,,,, P( ) ( 3 ) a r 0,, > ( 4 ) =, ( 5 ) ( a r ) = = a,x r { 0, },,, ( 6 ) { 0, },,,,, >, he obece funcon mnmzes he oal fxed and assgnmen coss. Consrans () guaranee ha, n eery me erod, each clen s fully assgned o exacly one ah; consrans (3) assure ha, n eery me erod, a clen can only be assgned o oen ahs; consrans (4) guaranee ha a facly can only be reoen a he begnnng of erod f has already been oen earler and s no n oeraon a he begnnng of erod ; consrans (5) mose ha a facly can only be oen once durng he lannng horzon; consrans (6) assure ha, n eery me erod, only one facly can be oen n each locaon. Consrans (5) and (6) need o be consdered exlcly only when here are negae fxed coss. Consrans (3) could be relaced by he aggregaed consrans: ( 7 ) ( a r ) x 0 = = P( ),, ( 3 ). Dual Problem and Comlemenary Condons Mullyng consrans (5) and (6) by -, and assocang dual arables wh consrans (), dual arables w wh consrans (3), dual arables u wh consrans (4), dual arables ρ 8

9 wh consrans (5 ) and dual arables π wh consrans (6 ), he dual roblem of DUMLP can be formulaed as D-DUMLP: D-DUMLP Max ρ π ( 8 ) subec o: w c,,, ( 9 ) w u = P ( ) = w = P( ) w ρ π u, u,, π 0 = π = FR FA,,, =, L, (0),, >, =, L, () ρ,,,, P( ) Consderng w = max{ 0, c } condensed formulaon s obaned: CD-DUMLP Max subec o: η, wh η = and η 0,,, P(), an equalen ρ η = P ( ) η = P ( ) u SA SR π { 0, c } FA u ρ max π,,, =, L, () = { 0, c } FR u max π,, >, =, L,, ρ, π 0,,, Le us defne: = FA = FR u = u { SA SR } ρ = π = π η = = P ( ) η = P ( ) = { 0, c } (3) max,,, =, L, (4) { 0, c } max,, >, =, L, ( 5 ) S = mn,,,, =, L, ( 6 ) Consderng he rmal roblem DUMLP and s dual CD-DUMLP, he followng comlemenary condons hold f n resence of omal rmal and dual soluons o he resece roblems (when here s no dualy ga): 9

10 ( a ) r x w = 0, = = = = = =,, P(), ( 7 ) a r u = 0,, > ( 8 ) = a ρ = 0, ( a ) r π = 0, = =, ( 9 ) ( 0 ) SA = 0,,, =,..., ( ) a SR = 0,,, =,..., ( ) r.3 Prmal-Dual Heursc he rmal-dual heursc ha has been deeloed bulds admssble rmal soluons based on admssble dual soluons o roblem CD-DUMLP, ryng o force he sasfacon of he comlemenary condons. If he heursc calculaes a ar of admssble and comlemenary rmal and dual soluons, hen he rmal omal soluon has been found. When hs s no ossble, he bes dual soluon found wll esablsh a ald lower lm o he omal alue of he rmal obece funcon. he heursc funconng scheme s as follows:. Inalsaon of dual arables;. Dual Ascen Procedure for dual arables 3. Prmal Procedure; ; 4. Dual Adusmen Procedure for dual arables ρ. If he dual soluon s changed go o ; 5. Reea he Prmal-Dual Adusmen Procedure for arables unl here s no mroemen n he dual obece funcon alue; 6. Dual Adusmen Procedure for dual arables ρ. If he dual soluon s changed go o ; 7. Dual Ascen Procedure for dual arables 8. Dual Descen Procedure for dual arables u. If he dual soluon s changed go o ; u. If he dual soluon s changed go o ; 9. Dual Adusmen Procedure for arables π. If he dual soluon s changed go o. hs heursc funconng scheme s equal o he funconng scheme resened n Das e al (004b). In fac, he only rocedures ha are dfferen from he ones descrbed n Das e al (004b) are he Dual Ascen Procedure and he Prmal-Dual Adusmen Procedure for arables Prmal Procedure. For hs reason, only hese hree rocedures are gong o be descrbed here., and he 0

11 .3. Dual Ascen Procedure for arables hs rocedure res o ncrease all dual arables, J, J J. When hs rocedure s execued n se of he rmal-dual heursc, hen J s equal o J. Oherwse, he se J s defned before hs rocedure s execued. hs rocedure s a sraghforward adaaon of he one descrbed n Van Roy and Erlenkoer (98). he only dfference s n he udang se of slacks SA and SR wh such ha more han one ah P( ), slacks c SA and and P( ). I s neresng o noe ha f c for SR, wh, wll be decremened more han once (hs s a comleely dfferen behaor, when comared wh he sngle-leel case (Erkenkoer, 978; Van Roy and Erlenkoer, 98). Dual Ascen Procedure for Varables. Consder any admssble nal soluon { } k each (, ), defne k(,) = mn{ k : c }. If q. J. (, ) (, ) ; ; δ 0 3. If (, ), hen go o 7. such ha c ( ) S k c (, ),,, 0, = hen (, ) k(, ) k.,,. For S 4. mn : (). If = 0, go o 7. η P c k ( ) 5. If, k (, ) > c hen c ; δ ; k(, ) k(, ). 6. For each facly, SR SR η and SA SA η,.. 7. If # J,(, ) (, ) ; q q q q. Go o If δ = go o. Else SOP. () P c () P c he calculaon of he η arameers can be done n seeral dfferen ways. One ery smle way s o consder η = n, where n s equal o he number of serces belongng o ah. Anoher calculaon rocedure s moaed by he analyss of se 4 of he dual ascen rocedure. I s sraghforward o conclude ha η should ake smaller alues for serces such ha he quoen beween ' mn{ S } S =,, and he number of ahs ncludng such ha c s smaller. hs rocedure s more me consumng han he frs, bu, n general, calculaes beer dual soluons.

12 Calculaon of η arameers. ;. If c hen go o 3. Else go o If n =, hen η, P(). Go o If P( ) hen go o 6, else go o 7., f c 6. e, n e, η 0, oherwse P() 7.. If > m, hen go o 8. Else go o 5. η 8. D η. Udae η,. D 9.. If > q hen so. Else go o. S mn. n Examle : Consder a roblem wh four clens, wo me erods and hree serces organzed n wo leels as deced n fgure. 3 Fgure : Clen; Serce. he admssble ahs are: =(,3), =(,3) and 3 =(3). ables and show he assgnmen coss for each me erod and able 3 shows he fxed (re) oenng coss. Varables are nalzed as showed n able 4. he dual ascen rocedure for arables begns by consderng = and =. For ah equal o, he rocedure calculaes 0 8 η = = and η = =.. Varable can be ncreased o 8. 8 For arable, he rocedure calculaes η = 0 56 and η = If. 3. was ncreased by 7 ( ) k, ( c ), hen slack SA would become equal o ha s less han zero, so s calculaed as beng equal o. No more dual arables are changed because slacks become as deced n able 5.

13 able able FA FA FA FR able FA FA able 4 able 5 3. For equal o and equal o, η = η = 0 5. If he alue (, ) k c = 4 was consdered, hen slack SA would be changed o 0.5 4, ha s less han zero. For hs reason s calculaed as beng equal o. Afer he dual ascen rocedure, he fnal resul s deced n ables 6 and FA FA FA FR able 6 able 7.3. Prmal-Dual Adusmen Procedure for arables he Prmal-Dual adusmen rocedure for arables deecs olaons of he comlemenary condons (7), and decreases he alues of some arables, ncreasng slacks and allowng oher arables o ncrease. hs rocedure res o reduce he number of comlemenary condons olaons and, smulaneously, mroe he alue of he dual obece funcon. 3

14 Comlemenary condon (7) wll be olaed f here exss a leas wo oen or arally oen ahs and such ha c > and c >. Noce ha sn necessary ha ahs are oen. hey only need o be arally oen o olae he comlemenary condons. Dmnshng arable wll ncrease all slacks SA and SR, wh, such ha > c, wh P(). If here are wo oher dual arables blocked exclusely due o slacks ha are ncreased, hen s ossble o mroe he dual obece funcon alue. I s neresng o noce ha he olaon may occur due o a sngle facly ha belongs o more han one ah (n he sngle case here had o be a leas wo dfferen facles noled). In he sngle-leel case, he dual ascen algorhm res frsly o ncrease arables blocked exclusely by slacks corresondng o a sngle facly (Das e al, 004a). he dual adusmen rocedure n he mul-leel case res frsly o ncrease all arables blocked exclusely due o slacks ha are gong o be ncreased (een f corresondng o more han one facly). Consder he followng defnons: I * = { (,, ): S = 0} I * = { : (,, ) I * and } I = { : facly s oen durng erod } I * = * { : (, ) wh and P( ) (,, ) I and c } { : I and P() such ha c } I = > { : such ha I and c } * {( ) * ( ) *, : I < < } {( ) I and, γ, I, I, γ or γ, : I I } { c : c } P = > J = c = max > Se I * corresonds o (,,), such ha SA and/or each clen, all oen or arally oen ahs such ha SR are equal o zero. Se P s greaer han he assgnmen cos ndcaes, for c. A olaon of he comlemenary condons (7) s deeced by he exsence of, a leas, one ar (,) such ha he number of elemens of P s greaer han one. he se I ndcaes, for each clen, all oerang facles durng erod ha belong o any ah such ha s greaer han he assgnmen cos c. hs means ha all slacks SA and SR,, wh I wll be ncreased wh he decrease n. he se * I corresonds o he se of all facles such ha here exss a leas one slack SA or SR,, blockng arable. Se J reresens all dual 4

15 arables ha can be ncreased wh he decrease of arable belongs o se ascen rocedure s called. Consder arables. I s ossble ha arable self J. Neerheless, hs arable won be consdered n se J a he frs me he dual organzed as a sequence of ars (,). Prmal-Dual Adusmen Procedure for Dual Varables. Inalze (, ) (, ) ; q ; δ 0.. If # hen go o 9. P 3. If J \{(, ) } 4. For each (, ) c. 5. J = J \{(, ) } =, hen go o 9.,, SA SA ( c ), SR SR ( c ), {( ) } () P > c. Execue he dual ascen rocedure for arable J = J,. Execue he dual ascen rocedure for arable J = J. Execue he dual ascen rocedure for arable. P() ; > c 6. If has been changed, go o. 7. Execue he rmal rocedure. 8. If here hae been mroemens n he dual or rmal obece funcon alue, hen δ 0. Else δ δ. 9. If he rmal obece funcon alue s equal o he dual obece funcon alue, or δ=δ max, or q # J hen so; else q q ; (, ) (, ) q, go o..3.3 Prmal Procedure he rmal rocedure consrucs rmal feasble soluons based on dual feasble soluons, ryng o force comlemenary condons o be sasfed. Consder he ses I *, I and I * defned n he reous secon, and also: I A = { (,, ) :.. a = }, I R = { (,, ) : r = }, P = {: s oen durng erod }. Ses I * and I are no necessarly equal, because he rmal rocedure wll always ry o oen he mnmum number of serces, guaraneeng ha all clens wll be assgned o one oen ah n eery me erod. Furhermore, s ofen necessary o nser n se I serces ha don belong o I *. Ses I A and I R are bul durng he rmal rocedure and deermne whch serces wll be (re) oened, when and for how long. Defnon 4: A ah s consdered essenal durng erod f here s a leas one clen ha has o be assgned o ah durng erod. hs haens f and only f J : c < c, ' P, '. ' 5

16 Pahs consdered as essenal are he frs o be consdered oen. o oen a ah a me erod all facles belongng o ha ah hae o be oen. hs s acheed by nserng all hose facles n se I. When oenng a ah, he rmal rocedure wll ofen olae comlemenary resrcons () and (), by ncludng n se I facles I *. Wheneer ah s oen, all oher ahs such ha f hen wll also be oen. Pahs no consdered essenal wll only be exlcly oen durng me erod f here are clens ha canno be assgned o already oen ahs. In hs case, he rocedure wll oen he ah ha corresonds o he smalles assgnmen cos. Ses I A and I R are buld based on ses I,. hese ses are bul usng exacly he same rocedures descrbed n Das e al (004a). Prmal Procedure. I A = I R =. I =,. Buld ses I * and I *. Num = 0;. For = unl, nclude n se P all ahs such ha c and < c, '. : ' Udae ses P,, ncludng n P all oen ahs P. 3. For each clen such ha c, P, nclude n se P ah such ha c c ' = mn c. Num Num. ' < 4. Include n se I all facles belongng o ah P,. 5. If Num = hen I * I and I, P,, go o. Else go o Buld ses I A and I R. Udae I and P. 7. For = unl, assgn each clen o ah P such ha c = mn { c } ' P. Calculae Z as beng he rmal obece funcon alue. 8. G = I A I R. 9. Choose arbrarly a arable a or r belongng o G and change s alue from one o zero. If he soluon remans admssble, recalculae he assgnmens of clens o oen ahs. Calculae Z as he rmal obece funcon alue of he new soluon. If Z < Z, hen remoe arable a (or r ) from se I A (or I R ) and se Z = Z. 0. Remoe arable a (or r ) from se G. If G =, go o. Else go o 9.. es comlemenary condons (9)-(). Ses 8-0 ry o mroe he rmal soluon calculaed by decreasng he number of facly locaon arables ha are consdered equal o one. hs has roed o decrease sgnfcanly he alue of he rmal obece funcon. hese ses could also be relaced by a dro heursc: from all arables equal o one, choose he one whch, when s alue s changed o zero, leads o he greaes mroe n he rmal obece funcon alue. Reea he rocess unl here s no mroemen n he obece funcon alue. Such a dro heursc was esed, bu he resuls obaned showed ha s much more me consumng and he alue of he rmal obece funcon 6

17 obaned s he same as wh he execuon of ses 8-0. hs s usfed by he fac ha he rmal rocedure consders more arables equal o one han he ones srcly needed (because he oenng of a ah consss n oenng a se of facles, ha mos of he mes oen mlcly oher ahs). hese arables wll always be consdered equal o zero, een f chosen arbrarly. Se res o change he rmal soluon n order o guaranee he sasfacon of comlemenary condons (8)-(0) ha are beng olaed. hs es s equal o he one already deeloed by he auhors o he sngle-leel case (Das e al, 004b). Consderng he dual soluon resened n ables 6 and 7, he rmal rocedure would oen ah, consderng facles and 3 oen from erod one o wo ( a = a ). hs rmal 3 = soluon has an obece funcon alue equal o 73, hus s he omal soluon. 3 Includng maxmum caacy consrans In almos all real suaons, a facly has an uer lm on he demand can sere. So, he ncluson of maxmum caacy resrcons n he roblem DUMLP s a naural exenson. Consder: Q = maxmum caacy of facly durng a me erod, x = fracon of cusomer s demand sered by ah, durng me erod. d = oal demand of cusomer durng me erod. he mul-leel, dynamc, caacaed locaon roblem (DCMLP) can be formulaed smly by ncludng n DUMLP consran (3) below and changng (7) o (7 ): ( a r ) Q = = { 0, } { 0, } d x 0,, ( 3 ) P() a,,, r,, >, ( 7 ) x 0,,, he addonal se of resrcons (3) esablsh an uer lm on he oal flow ha reaches an oen facly n each me erod. o deelo a rmal-dual heursc, s necessary o obsere he changes hese addonal resrcons brng o he dual condensed roblem formulaon. Assocang dual arables λ o resrcons (3), he condensed dual roblem formulaon becomes: 7

18 CD-DCMLP Max ρ π subec o: max 0, c d η = P( ) ηmax, c d = P( ) u λ u = FA ρ π Q = = λ,,, =, L, ( 4 ) 0 λ FR u π Q λ, = =, ρ, π, λ 0,,,,, =, L, ( 5 ) A decrease or ncrease n dual arable λ wll nfluence all slacks S ', wh, such ha P() P( ), I (n he calculaon of S ', he sum oer all ahs P( ) wll consder all alues of λ such ha belongs o ). he rmal-dual heursc s funconng scheme resened for he uncaacaed case remans ald for he caacaed case, wh wo addonal ses: 0. Dual Ascen Procedure for dual arables λ. If he dual soluon s changed, go o.. Dual Descen Procedure for dual arables λ. If he dual soluon s changed, go o. he dual and descen rocedures for arables λ wll be descrbed n he nex wo secons. he deelomen of hese rocedures followed he work of Gugnard and Selberg (979), Saldanha da Gama (00) and Das e al (004b). he rmal rocedure wll hae o be changed ( s necessary o es he sasfacon of he addonal caacy resrcons). he dual ascen and Prmal-Dual adusmen rocedures for arables he assgnmen coss equal o c d reman ald beng only necessary o consder λ. All he remanng rocedures are no changed. 3. Dual Ascen Procedure for Varables λ If arable λ s ncreased, hen he lef hand sde of consrans (4) and (5) wll dmnsh. he maxmum change ha should be consdered s such ha all hose alues became less han or 8

19 equal o zero. Consderng ha arable λ s ncreased by δ, hen δ should be less han or equal o () such ha: () c d λ = max J P () d ( 6 ) Consder he followng defnons: (,', δ ) = (, ) : J P( ) P( ' ) c d JP d c d JP(,', δ ) = (, ) : J P( ) P( ' ) d λ λ δ, ( 7 ) > δ. ( 8 ) Prooson : If dual arable λ s ncreased by δ ]0, ()], hen slacks I such ha P() P( ), wll be changed by: SA' and SR ',, Ω(δ,, ) = η ' max 0, c d λ η' d δ E ( 9) (,) JP(,', δ ) (,) JP (,', δ ) where: Q, f = ' E = δ. 0, oherwse hs rooson follows smlar resuls ha can be found n Gugnard and Selberg (979). hs rooson moaes he dual ascen rocedure ha s now descrbed. Dual Ascen Procedure for Varables δ 0; δ. 4. δ max max 0, c d λ. J () d P max 0, c d λ < δ' d λ 5. Calculae ses JP(,,δ) and JP (,,δ) as n (7) and (8), and Ω(,,δ) as n (9), I, wh P() P( ). 6. If Ω(δ,, ) < 0, I, wh P() P( ), hen go o. Else go o If I, wh P() P( ), such ha SA ' Ω(δ,, ) < 0 or SR ' Ω(δ,, ) < 0,, hen go o 8. Else go o 9. 9

20 8. If δ = 0, hen go o. Else δ δ and go o λ λ δ; SA ' SA ' Ω(δ,, ) and SR' SR ' Ω(δ,, ),, I such ha P() P( ). 0. Execue he dual ascen rocedure for arables, wh J = J... If > M hen go o. Else go o 3... If > hen so. Else go o. Examle : Consder a roblem wh sx clens, hree serces organzed as n fgure, and wo me erods. he clens hae a known demand ha s equal o 56,, 76, 58, 9 and 57 n me erod and equal o 59,, 78, 6, 8 and 58 n me erod, for =,,6. Serces hae maxmum caaces equal o 49, 6 and 45. Imagne ha he frs me he dual ascen rocedure for arables λ s execued arables are equal o zero and slacks hae alues resened n able 9. ake on he alues resened n able 8, all oher dual arables SA SA able 8 SA SR able 9 Consderng arable λ : δ ; P()={}. he alue of δ s calculaed as: δ Ω max max max { 0, }, max{ 0474, 034}, max{ 0, }, 9 { 0, }, max{ 0, 53 53}, max{ 0, } = ( 8. 69,, ) = ( ) η ( ) η ( ) ( ) η η Consderng η = 0. 3, η = 0. 76, η 3 = and η 6 = 0. 87, hen ( 8. 69,, ) = Ω. Varable λ canno be ncreased by δ because he rocedure deecs ha slack SA would become less han zero. he alue of δ s changed o 8.69 and he rocedure s reeaed, obanng δ equal o 0. 0

21 Ω ( 0,, ) = ( ) 0. 3 ( ) ( ) = hs change s sll no admssble, so δ s changed o 0 and δ wll become equal o. Ω (,, ) = ( ) ( ) = All slacks P() P() = S wll connue greaer han or equal o zero. Slacks S are no changed because ( 3,, ) = ( ) ( ) Ω = hs means ha slacks S ,, are gong o be ncreased. he udaed slacks alues, are shown n able 0. he ncrease n arable λ allows he mroemen n he dual obece funcon alue, because arable s ncreased o able shows he slacks alues afer he execuon of he dual ascen rocedure for arables. SA SA SA SR SA SA SA able 0 able 3. Dual Descen Procedure for Varables λ SR If dual arable λ s decreased all alues c d λ, wh P(), wll be ncreased. Consder a decrease n λ such ha all hose alues ha are less han zero reman ha way. hs means ha λ can be decreased by δ such ha: 0 < δ mn c d λ J, P () d c d < λ 0 ( 30 ) changed. Once agan all slacks SA' and SR ',, I such ha P() P( ), wll be Prooson : If arable λ s decreased by a alue δ n he neral defned by (30), hen SA' and SR ',, I such ha P() P( ), wll be changed by:

22 (,',) δ = δ E d η', ( 3 ) J P () P( ' ) λ 0 Φ c d where Q, f = ' E =. 0, oherwse hs rooson follows smlar resuls ha can be found n Gugnard and Selberg, (979). Prooson moaes he followng dual descen rocedure. Dual Descen Procedure for Varables λ δ mn λ, mn c d λ. J, P () d c d λ < 0 4. Calculae Φ(δ,,) as n (3), I, wh P() P( ). 5. If Φ(δ,,) < 0, I, wh P() P( ), hen go o 0. Else go o If I, wh P() P( ), such ha SA ' Φ(δ,,) < 0 or SR ' Φ(δ,,) < 0,, hen go o 7. Else go o 8. S ' 7. δ mn Φ( ). If δ = 0 hen go o 0. Else go o 8. δ,', δ 8. λ S ' Φ ( δ,', ) λ - δ; < 0 SA SA Φ(δ,,) and ' ' SR' SR ' P( ). 9. Execue dual ascen rocedure for arables, wh J = J. 0.. If > M hen go o. Else go o 3... If > hen so. Else go o. Φ (δ,,),, I such ha P() Examle 3: Consderng examle and arable λ equal o. hen, for equal o he rocedure calculaes: = d. c d 0 η and Φ (,, ) = = 0. 7 Consderng equal o 3 ( s no needed o consder equal o because P() P() = ) he rocedure calculaes Φ (, 3, ) = hs means ha slacks S ',, wll ncrease for

23 equal o and decrease for equal o. Slack equal o zero n se 7 of he rocedure, so he dual arable s no changed. 3.3 Prmal Procedure SA 3 s equal o zero, whch means ha δ wll be he rmal rocedure ha bulds admssble soluons o DCMLP s ery smlar o he one descrbe for DUMLP. Afer buldng ses I A and I R, s necessary o es he sasfacon of he maxmum caacy resrcons. he assgnmen of cusomers o ahs s acheed by solng ranshmen roblems. If he ranshmen roblem s no feasble for some me erod hen wll be necessary o oen more ahs (noce ha s no ossble o esablsh a necessary and suffcen condon of he form oal caacy of oen facles greaer han or equal o clen s oal demand o es he admssbly of a gen soluon, as haens n he sngle leel case). o choose wha ah o oen, he rocedure calculaes he mnmum cos of oenng a ah by consderng he coss of oenng all serces belongng o ha are no oeraonal durng me erod. Consder: F = mnmum cos of oenng facly I durng erod ; H = mnmum cos of oenng ah P durng erod ;, f P e =,, ; 0, oherwse n = e. P () Prmal Procedure. Execue se -6 of he DUMLP s rmal rocedure Sole a ranshmen roblem consderng as sources he se J of clens (wh sules d ), as desnaons he se of facles I belongng o leel K (wh demands Q ), and as ranshmen ons he se of facles I belongng o leels o K- (wh demand and suly equal o Q ). If he roblem s no feasble go o 4. Else go o If P, hen so. he roblem s nfeasble. Else go o Calculae H = F and I Q Ca = mn, P. n 6. Choose ah such ha H ' Ca ' = H mn Ca P. 7. Oen ah, ncludng n se P and n se I,. Go o 3. 3

24 8.. If > hen go o 9. Else go o Calculae Z as he curren rmal obece funcon alue. 0. Execue ses 8-0 of DUMLP s rmal rocedure. he mnmum cos of oenng a facly durng erod (F ) s calculaed as descrbed n Das e al (004b). Calculaon of Ca res o llusrae he maxmum caacy of ah. he maxmum caacy of ah s deermned by he facly wh mnmum caacy. If facly belongs o more han one ah, s caacy s consdered o be equally dded by all oen ahs. 4 Pahs Whou Flow Conseraon Consder ha facles belongng o leels,,k- hae a arameer θ such ha f a flow of d uns reaches hen hs facly wll only ass o he nex facly a flow d θ (see fgure 3). hs s a generalsaon of DCMLP, where θ s equal o one for all facles. d Fgure 3 d θ here are seeral examles where one can fnd hs knd of behaor n nermedae facles. In sold wase reamen sysems, ncneraors recee a ceran amoun of wase o burn, and only a small ar of (n he form of ashes) has o be laced n landflls. In a healh facly srucure, only a small grou of clens ha are sered n local healh cenres hae o be conduced o local hosals. he consderaon of arameers θ assocaed wh nermedae facles comlcaes exremely he roblem, secally he deermnaon of omal assgnmens of clens o ahs (when he locaon arables are already fxed o one or zero). In fac, hs assgnmen can no longer be calculaed omally hrough he resoluon of ransshmen roblems. In he nex aragrahs, he rmal and condensed-dual formulaons are resened, and he rmal-dual heursc deeloed s descrbed. he comlemenary condons are no saed because hey are ery smlar o (7)-(). he only dfference beween hs roblem and DCMLP les n he maxmum caacy resrcons. Consder he followng defnon: Defnon 5: Le and be wo facles and a ah such ha and. hen f facly aears before facly n he ordered se. < ' f and only 4

25 he mul-leel, dynamc caacaed locaon roblem whou flow conseraon on he nermedae facles (DCMLP) can be formulaed as DCMLP wh consrans (3) below nsead of consrans (3). Q ( a r ) dx θ' 0,, ( 3 ) = = P() ' < Consderng he same dual arables as n CD-DCMLP, he condensed-dual roblem becomes: CD-DCMLP Max subec o: ρ π max 0, c d θ' λ FA u ρ π Q λ ' < = = =, η = P( ),, =, L, ( 33 ) max 0, c d θ' λ FR u π Q λ, ' < = = η = P( ) u, ρ, π λ 0,,,,,, =, L, ( 34 ) he rmal-dual heursc funconng scheme s he same resened n secon 3, wh an addonal se:. Calculae he omal assgnmen of clens o ahs consderng ses I and P ha corresond o he bes rmal soluon calculaed hus far. Some rocedures had o be slghly changed. Consder ' ' < d = d θ, P(),. he dual ascen and rmal-dual adusmen rocedures for arables do no need o be changed, beng only necessary o consder assgnmen coss c = c d ' ' λ,, P(),. he changes n dual ascen and descen rocedures for arables λ are a drec consequence of d defnon. In he dual ascen rocedure he defnons (6) o (8) should be relaced by (35) o (37): 5

26 () c d' λ = max, ( 35 ) J P () d' ' (,', δ ) = (, ) : J P( ) P( ' ) c d JP d' c d ' ' JP(,', δ ) = (, ) : J P( ) P( ' ) d' λ λ δ, ( 36 ) > δ, ( 37 ) In hs case Ω(δ,, ) s gen by: Ω(δ,, ) = η ' max 0, c d' λ η' d' δ E ( 38) (,) JP(,', δ ) (,) JP (,', δ ) where: Q, f = ' E = δ. 0, oherwse 4. δ Se 4 of he dual ascen rocedure for arables λ should be relaced by: 0 max d max max 0, c d' λ J () d P ' ', c d ' λ < δ' In he dual descen rocedure for arables λ, he followng changes should be consdered: δ = δ E d ' η ', ( 39 ) J P () P( ' ) c d ' λ 0 (,',) Φ where: 0 < δ mn c d () ' λ. ( 40 ) J, P d ' c d λ < 0 ' 6

27 Se 3 should be changed o: 3. δ mn λ, mn () J, P ' c d ' λ < 0 d' c d ' λ. he rmal rocedure s ery smlar o he one already resened n secon 3. Se 5 s changed o: 5. Calculae H = F and I Ca Q = mn, P. n θ' ' < hs change res o accoun no only for he number of ahs o whch facly belongs, bu also for he relae oson of facly n hose ahs. he assgnmen roblem canno be soled as a ranshmen roblem (se 3 of he rmal rocedure). Afer calculang ses I and P, he followng lnear rogram should be soled o fnd he omal assgnmens of clens o ahs, for each me erod : AP() Mn ( 4 ) subec o: P c x x =, ( 4 ) x d θ ' Q, I ( 43 ) P() P ' < x 0,,, Solng hs roblem omally (usng a general soler) n each execuon of he rmal rocedure (and for all me erods) s ery me consumng. So, a heursc was deeloed o fnd feasble soluons o AP(). hs heursc rocedure s used n se 3 of he rmal rocedure (nsead of he resoluon of he ransshmen roblem). Problem AP(),, s only soled omally a se of he rmal-dual heursc (usng a general soler), consderng he locaon arables ha corresond o he bes rmal soluon calculaed by he heursc unl ha momen. 7

28 Heursc Procedure for he resoluon of roblem AP() Q. D d,. Q θ ' ' <,, P(), Ca mn{ Q }, P. x 0,,.. If Ca = 0, P and such ha D > 0, hen so: he roblem s mossble. c mn c and 3. For each clen J, wh D >0, calculae { } c { c : c c } mn P Ca > 0 4. Choose clen and ah such ha c c = max{ c ' c ' } ' J 5. δ mn{ D, }, x x δ Q Ca Q, δ 6., P,, such ha, Q 7. If D = 0,, hen so. Else, go o... P Ca > 0 Q,, Ca δ, D D δ. Q δ θ ' ' < θ ' ' < and mn{ Ca, Q } Ca. Calculaon of Q exresses he maxmum flow ha can be assgned o ah due o he maxmum caacy of facly. he mnmum of hese caaces corresonds o he maxmum caacy of ah desgnaed by Ca. Obseraon : Consder wo ses (S and S ) of locaon arables a and r. Consder wo admssble soluons o DCML consruced by consderng equal o one all locaon arables belongng o S and S, resecely, beng all he oher locaon arables equal o zero. Le us consder ha Z * <Z *, where Z * * and Z reresen he omal obece funcon alue corresondng o soluon S and S, resecely. If Z and Z reresen he obece funcon alue obaned by consderng soluons S and S, resecely, and solng he assgnmen roblems AP() heurscally, hen Z Z * and Z Z *, bu s no ossble o guaranee ha Z < Z. hs has an moran consequence: s ossble ha he rmal-dual heursc fnds soluons beer han he fnal soluon resened, bu ha he heursc s no able o denfy as beng beer. hs shor come canno be soled, unless he omal soluons o AP() are calculaed for eery me erod, wheneer he rmal rocedure s execued. As has already been sad, ha would be rohbe n erms of execuon mes. Examle 4: Consder a roblem wh four clens and fe oenal serces such ha serce and are n leel, serces 3 and 4 are n leel and serce 5 s n leel 3. here are fe admssble ahs: 8

29 =(,3,5), =(,4,5), 3 =(,4,5), 4 =(,5), 5 =(4,5). A erod he clens demands are as follows: d =5; d =0; d 3 =8; d 4 =. Serces o 5 hae maxmum caaces equal o 0, 50, 80, 70 and 00, resecely, and θ alues equal o 0.8, 0.5, 0.5, 0.. Se : Q = 0 ; Q = = ; Q = = ; { Q,Q, } 0 Ca ; = mn 3 Q5 = Q = 0 ; Q 4 = =. ; Q 00 5 = = ; { Q,Q, } 0 Ca ; = mn 4 Q5 = Q 3 = 50 ; Q = = ; Q = =.. ; { Q,Q, } 50 Ca ; 3 = mn 3 43 Q53 = Q = 50 ; = = 0. 5 Ca ; 4 = mn Q4, Q54 = Q ; { } Q 45 = 70; 55 = = Q ; = mn{ Q, Q } 70 Ca = Se 3: hs nformaon s summarzed n able Ca able c c c able 3 Se 4: Choose clen equal o 3 and ah equal o. Se 5: δ mn{8, 0}=8. D 3 0; Values c Q and Ca are changed accordng wh able 4. 9

30 = 0-8= = =. 0 = = = = = 97. = Ca able 4 Se 4: Choose clen equal o and ah equal o. Se 5: δ mn{0, }=; D 8; Values Se 3: Q and Ca are changed accordng wh able 5 (he bold alues are he ones ha are changed) Ca able 5 c c c c able 6 Se 4: Choose clen equal o and ah equal o 3. Ses 5 and 6 : δ mn{5, 50}=5 D 0; Values able 7 (he bold alues are he ones ha are changed). Se 4: Choose clen equal o and ah equal o 5. Ses 5 and 6: δ mn{8, 46.5}=8. D 0; Values able 8. Q and Ca are changed accordng wh Q and Ca are changed accordng wh 30

31 Se 4: Choose clen equal o 4 and ah equal o 3 Ses 5 and 6: δ mn{, 35}=; D 4 0; Se 7: Values D are equal o zero, for all, whch means ha an admssble soluon has been found (he soluon found s llusraed n fgure 4): P I Ca Ca able 7 able Fgure 4 Problem DCMLP can be generalzed consderng a new se of consrans mosng mnmum caacy resrcons: Q' ( a r ) d x θ' 0 = = P (),, ( 44 ) ' ' < whereq' reresens he mnmum flow ha has o reach an oen facly, durng he me erods hs facly s oerang. Assocang dual arables β wh consrans (44), he condensed dual formulaon s slghly changed: CD-DCMLP Max subec o: ρ π 3

32 max 0, c d θ' λ β ' <,,, =, L, (45) FA u ρ π Q λ β Q' η = P( ) = = η max 0, c d θ' λ β = P( ) ' <,,, =, L, (46 ) FR u π Q λ β = = Q' = u =, ρ, π, λ, β 0,,, = As can be seen, he behaor of dual arables β and λ s symmerc, so he dual ascen rocedure descrbed for dual arables λ consues he dual descen rocedure for β, and he dual descen rocedure for dual arables λ consues he dual ascen rocedure for arables β. he rmal rocedure has o be changed, because s necessary o sasfy a new se of consrans. Neerheless, he new rmal rocedure consruced s, n essence, smlar o he one reously descrbed. he only dfference consss n he heursc rocedure o sole he assgnmen roblem (roblem AP() has also a new se of mnmum caacy consrans), and also n some addonal ses ha ry o close or nerchange oen and closed facles, when an admssble soluon s no found by oenng new facles (usng he rocedure already descrbed). he omal assgnmen of clens o facles, consderng he locaon arables fxed o one or zero, can be found by solng he followng lnear roblem: AP() Mn subec o: P Q' c x x =, ( 4 ) x d P P ' < () θ Q, I ( 47 ) ' x 0,,, 3

33 Heursc o fnd an admssble soluon o AP(). Calculae an admssble soluon o AP(). If he calculaed soluon s admssble o AP() so. Else go o.. 0,, ; Camn Q' and Camax Q, I ; D d,. x 3. If Ca 0, I, hen go o If D = 0, and I such ha Camn >0, hen so: he heursc canno fnd an admssble soluon. Else go o 5. Camn 5. Choose facly, wh Camn > 0, such ha Camn = max '. Camax ' I Camax' 6. If P P() such ha Camn > 0, hen go o 7. Else go o 8. c c ' ' 7. Choose ah P P() and clen such ha = mn. Go o 9. θ ' ' P() P θ ' ' < Camn > 0, ' ' < ' ':D ' > 0 c c ' ' 8. Choose ah P P() and clen such ha = mn. θ ' ' P() P θ ' ' < ' :D ' > 0 ' < ' 9. δ Camax ' Camn mn D,mn, ' θ'' θ ; x δ x ; D D δ; ' '' < ' ' < Camn Camn δ θ and Camax Camax δ θ,. Go o 3. ' ' < ' ' < 0. If such ha D > 0, hen sole heurscally roblem AP() consderng Q equal o Camax, I and d equal o D,. Else so. If an admssble soluon s no found n se of hs heursc, hen res o assgn flow o ahs n order o sasfy he mnmum caacy consrans of oen facles. Se 5 of hs heursc chooses facly such ha he relaon beween s maxmum and mnmum caaces s more dffcul n erms of sasfyng boh resrcons. he heursc wll ry frs o assgn flow o ahs P() such ha no facly belongng o ah has s mnmum caacy resrcon sasfed (se 6). When hs s no ossble, he heursc consders all oen ahs P() (se 7). In ses 6 or 7 he heursc chooses a clen and an oen ah ha corresond o he mnmum cos er un ha reaches facly. he ncrease n he flow ha s assgned o ah s calculaed as beng he mnmum beween he remanng demand of clen and he necessary flow o sasfy he mnmum caacy resrcon of facly, guaraneeng ha he maxmum caacy resrcons for all oher facles belongng o ah reman admssble. When he heursc reaches se 0, all mnmum 33

34 caacy resrcons for oen facles hae been sasfed. If here are sll clens hang a remanng demand greaer han zero, hen wll be necessary o assgn hese demands o ahs usng he already descrbed heursc for roblem AP(). he descrbed heursc canno guaranee he consrucon of an admssble soluon o AP() wheneer he roblem s feasble. I s always ossble o check f he roblem AP() s nfeasble by usng a general soler (whch wll be more me consumng bu wll fnd ou, for sure, f he roblem s or s no nfeasble, when he heursc s unable o fnd an admssble soluon). Examle 5 Consder he roblem descrbed n examle 4, wh facles hang he followng mnmum caaces: 5, 0, 0, 60, 5. he mnmum caacy resrcon s no sasfed for facly 4, so he soluon found n se of he heursc s no admssble. Se : Camn 5; Camn 0; Camn 3 0; Camn 4 60; Camn 5 0. Camax 0; Camax 50; Camax 3 80; Camax 4 70; Camax D 5; D 0; D 3 8; D 4. Se 5: Choose facly 4 because = max,,, Se 7: Choose ah, 3 or 5 and a clen accordng o he calculaons resened n able 9. he heursc wll choose ah and clen able Se 9: δ mn 8, 0,,, = 8 ; x 3 8 ; D 3 0; Camn 5 8= 3; Camn =45.60; Camn =3.56. Camax ; Camax ; Camax Se 5: Choose facly 4 because = max,, Se 7: he heursc wll choose ah 5 and clen 4. 34

35 Se 9: δ mn, ,, = ; x ; D 4 0; Camn ; Camn 5.36; Camax ; Camax Se 5: Choose facly 4 because = max,, Se 7: he heursc wll choose arbrarly beween clen and ah 3 or clen and ah 5 (he heursc wll neer choose clen and ah because ah has facles wh he mnmum caacy resrcon sasfed). Consder ha he heursc chooses clen and ah Se 9: δ mn 5, 50,,, = 5 ; x 3 5 ; D 0; Camn 5; Camn 4 6.0; Camn 5 0.6; Camax 35; Camax 4 6.0; Camax Se 5: Choose facly 4 because = max, Se 7: he heursc chooses ah 5, because s he only ah ha s consued by facles whose mnmum caacy resrcon s no ye sasfed, and chooses clen Se 9: δ mn 0, 6.,, 6. 0 = 6. 0 ; x ; D 3.9; 0. Camn 4 0; Camn 5 ; Camax 4 0; Camax Se 0: he curren soluon s already admssble because Camn 0,. As D > 0, hen wll be necessary o use he heursc o sole roblem AP(). If he mnmum caacy resrcons were consdered n roblem DCMLP (wh θ equal o one, for all facles) hen he assgnmen roblem AP() would be a ransshmen roblem, wh each nermedae facly corresondng o wo ransshmen ons: one hang demand and suly equal o s mnmum caacy and he oher hang demand and suly equal o s maxmum caacy mnus s mnmum caacy. Consder: f = smalles cos ncurred by closng I durng erod. 35

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

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas)

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas) Lecure 8: The Lalace Transform (See Secons 88- and 47 n Boas) Recall ha our bg-cure goal s he analyss of he dfferenal equaon, ax bx cx F, where we emloy varous exansons for he drvng funcon F deendng on

More information

PHYS 705: Classical Mechanics. Canonical Transformation

PHYS 705: Classical Mechanics. Canonical Transformation PHYS 705: Classcal Mechancs Canoncal Transformaon Canoncal Varables and Hamlonan Formalsm As we have seen, n he Hamlonan Formulaon of Mechancs,, are ndeenden varables n hase sace on eual foong The Hamlon

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

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

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

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

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

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

Chapter Lagrangian Interpolation

Chapter Lagrangian Interpolation Chaper 5.4 agrangan Inerpolaon Afer readng hs chaper you should be able o:. dere agrangan mehod of nerpolaon. sole problems usng agrangan mehod of nerpolaon and. use agrangan nerpolans o fnd deraes and

More information

Outline. Energy-Efficient Target Coverage in Wireless Sensor Networks. Sensor Node. Introduction. Characteristics of WSN

Outline. Energy-Efficient Target Coverage in Wireless Sensor Networks. Sensor Node. Introduction. Characteristics of WSN Ener-Effcen Tare Coverae n Wreless Sensor Newors Presened b M Trà Tá -4-4 Inroducon Bacround Relaed Wor Our Proosal Oulne Maxmum Se Covers (MSC) Problem MSC Problem s NP-Comlee MSC Heursc Concluson Sensor

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

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

2/20/2013. EE 101 Midterm 2 Review

2/20/2013. EE 101 Midterm 2 Review //3 EE Mderm eew //3 Volage-mplfer Model The npu ressance s he equalen ressance see when lookng no he npu ermnals of he amplfer. o s he oupu ressance. I causes he oupu olage o decrease as he load ressance

More information

NPTEL Project. Econometric Modelling. Module23: Granger Causality Test. Lecture35: Granger Causality Test. Vinod Gupta School of Management

NPTEL Project. Econometric Modelling. Module23: Granger Causality Test. Lecture35: Granger Causality Test. Vinod Gupta School of Management P age NPTEL Proec Economerc Modellng Vnod Gua School of Managemen Module23: Granger Causaly Tes Lecure35: Granger Causaly Tes Rudra P. Pradhan Vnod Gua School of Managemen Indan Insue of Technology Kharagur,

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

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

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

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

A Cell Decomposition Approach to Online Evasive Path Planning and the Video Game Ms. Pac-Man

A Cell Decomposition Approach to Online Evasive Path Planning and the Video Game Ms. Pac-Man Cell Decomoson roach o Onlne Evasve Pah Plannng and he Vdeo ame Ms. Pac-Man reg Foderaro Vram Raju Slva Ferrar Laboraory for Inellgen Sysems and Conrols LISC Dearmen of Mechancal Engneerng and Maerals

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

Motion in Two Dimensions

Motion in Two Dimensions Phys 1 Chaper 4 Moon n Two Dmensons adzyubenko@csub.edu hp://www.csub.edu/~adzyubenko 005, 014 A. Dzyubenko 004 Brooks/Cole 1 Dsplacemen as a Vecor The poson of an objec s descrbed by s poson ecor, r The

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

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

Energy Storage Devices

Energy Storage Devices Energy Sorage Deces Objece of Lecure Descrbe he consrucon of a capacor and how charge s sored. Inroduce seeral ypes of capacors Dscuss he elecrcal properes of a capacor The relaonshp beween charge, olage,

More information

An ant colony optimization solution to the integrated generation and transmission maintenance scheduling problem

An ant colony optimization solution to the integrated generation and transmission maintenance scheduling problem JOURNAL OF OTOELECTRONICS AND ADVANCED MATERIALS Vol. 0, No. 5, May 008,. 46-50 An an colony omzaon soluon o he negraed generaon and ransmsson manenance schedulng roblem. S. GEORGILAKIS *,. G. VERNADOS

More information

Testing a new idea to solve the P = NP problem with mathematical induction

Testing a new idea to solve the P = NP problem with mathematical induction Tesng a new dea o solve he P = NP problem wh mahemacal nducon Bacground P and NP are wo classes (ses) of languages n Compuer Scence An open problem s wheher P = NP Ths paper ess a new dea o compare he

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

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

Online Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading

Online Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading Onlne Supplemen for Dynamc Mul-Technology Producon-Invenory Problem wh Emssons Tradng by We Zhang Zhongsheng Hua Yu Xa and Baofeng Huo Proof of Lemma For any ( qr ) Θ s easy o verfy ha he lnear programmng

More information

A New Generalized Gronwall-Bellman Type Inequality

A New Generalized Gronwall-Bellman Type Inequality 22 Inernaonal Conference on Image, Vson and Comung (ICIVC 22) IPCSIT vol. 5 (22) (22) IACSIT Press, Sngaore DOI:.7763/IPCSIT.22.V5.46 A New Generalzed Gronwall-Bellman Tye Ineualy Qnghua Feng School 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

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

[ ] 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

Dual Approximate Dynamic Programming for Large Scale Hydro Valleys

Dual Approximate Dynamic Programming for Large Scale Hydro Valleys Dual Approxmae Dynamc Programmng for Large Scale Hydro Valleys Perre Carpener and Jean-Phlppe Chanceler 1 ENSTA ParsTech and ENPC ParsTech CMM Workshop, January 2016 1 Jon work wh J.-C. Alas, suppored

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

. 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

Clustering (Bishop ch 9)

Clustering (Bishop ch 9) Cluserng (Bshop ch 9) Reference: Daa Mnng by Margare Dunham (a slde source) 1 Cluserng Cluserng s unsupervsed learnng, here are no class labels Wan o fnd groups of smlar nsances Ofen use a dsance measure

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

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

Pavel Azizurovich Rahman Ufa State Petroleum Technological University, Kosmonavtov St., 1, Ufa, Russian Federation

Pavel Azizurovich Rahman Ufa State Petroleum Technological University, Kosmonavtov St., 1, Ufa, Russian Federation VOL., NO. 5, MARCH 8 ISSN 89-668 ARN Journal of Engneerng and Aled Scences 6-8 Asan Research ublshng Nework ARN. All rghs reserved. www.arnjournals.com A CALCULATION METHOD FOR ESTIMATION OF THE MEAN TIME

More information

Mechanics Physics 151

Mechanics Physics 151 Mechancs Physcs 5 Lecure 0 Canoncal Transformaons (Chaper 9) Wha We Dd Las Tme Hamlon s Prncple n he Hamlonan formalsm Dervaon was smple δi δ Addonal end-pon consrans pq H( q, p, ) d 0 δ q ( ) δq ( ) δ

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

Reactive Methods to Solve the Berth AllocationProblem with Stochastic Arrival and Handling Times

Reactive Methods to Solve the Berth AllocationProblem with Stochastic Arrival and Handling Times Reacve Mehods o Solve he Berh AllocaonProblem wh Sochasc Arrval and Handlng Tmes Nsh Umang* Mchel Berlare* * TRANSP-OR, Ecole Polyechnque Fédérale de Lausanne Frs Workshop on Large Scale Opmzaon November

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

Beyond Balanced Growth : Some Further Results

Beyond Balanced Growth : Some Further Results eyond alanced Growh : Some Furher Resuls by Dens Sec and Helmu Wagner Dscusson Paer o. 49 ay 27 Dskussonsberäge der Fakulä für Wrschafswssenschaf der FernUnversä n Hagen Herausgegeben vom Dekan der Fakulä

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

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

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

APOC #232 Capacity Planning for Fault-Tolerant All-Optical Network

APOC #232 Capacity Planning for Fault-Tolerant All-Optical Network APOC #232 Capacy Plannng for Faul-Toleran All-Opcal Nework Mchael Kwok-Shng Ho and Kwok-wa Cheung Deparmen of Informaon ngneerng The Chnese Unversy of Hong Kong Shan, N.T., Hong Kong SAR, Chna -mal: kwcheung@e.cuhk.edu.hk

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

Political Economy of Institutions and Development: Problem Set 2 Due Date: Thursday, March 15, 2019.

Political Economy of Institutions and Development: Problem Set 2 Due Date: Thursday, March 15, 2019. Polcal Economy of Insuons and Developmen: 14.773 Problem Se 2 Due Dae: Thursday, March 15, 2019. Please answer Quesons 1, 2 and 3. Queson 1 Consder an nfne-horzon dynamc game beween wo groups, an ele and

More information

Simulation and optimization of supply chains: alternative or complementary approaches?

Simulation and optimization of supply chains: alternative or complementary approaches? Smulaon and omzaon of suly chans: alernae or comlemenary aroaches? hrsan Almeder *, Margareha Preusser, Rchard F. Harl Unersy of Venna, Brünnersr. 72, A-1210 Venna, Ausra {chrsan.almeder, margareha.reusser,

More information

EEL 6266 Power System Operation and Control. Chapter 5 Unit Commitment

EEL 6266 Power System Operation and Control. Chapter 5 Unit Commitment EEL 6266 Power Sysem Operaon and Conrol Chaper 5 Un Commmen Dynamc programmng chef advanage over enumeraon schemes s he reducon n he dmensonaly of he problem n a src prory order scheme, here are only N

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

Computational results on new staff scheduling benchmark instances

Computational results on new staff scheduling benchmark instances TECHNICAL REPORT Compuaonal resuls on new saff shedulng enhmark nsanes Tm Curos Rong Qu ASAP Researh Group Shool of Compuer Sene Unersy of Nongham NG8 1BB Nongham UK Frs pulshed onlne: 19-Sep-2014 las

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

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

Pattern Classification (III) & Pattern Verification

Pattern Classification (III) & Pattern Verification Preare by Prof. Hu Jang CSE638 --4 CSE638 3. Seech & Language Processng o.5 Paern Classfcaon III & Paern Verfcaon Prof. Hu Jang Dearmen of Comuer Scence an Engneerng York Unversy Moel Parameer Esmaon Maxmum

More information

THERMODYNAMICS 1. The First Law and Other Basic Concepts (part 2)

THERMODYNAMICS 1. The First Law and Other Basic Concepts (part 2) Company LOGO THERMODYNAMICS The Frs Law and Oher Basc Conceps (par ) Deparmen of Chemcal Engneerng, Semarang Sae Unversy Dhon Harano S.T., M.T., M.Sc. Have you ever cooked? Equlbrum Equlbrum (con.) Equlbrum

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

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that THEORETICAL AUTOCORRELATIONS Cov( y, y ) E( y E( y))( y E( y)) ρ = = Var( y) E( y E( y)) =,, L ρ = and Cov( y, y ) s ofen denoed by whle Var( y ) f ofen denoed by γ. Noe ha γ = γ and ρ = ρ and because

More information

THE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS

THE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS THE PREICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS INTROUCTION The wo dmensonal paral dfferenal equaons of second order can be used for he smulaon of compeve envronmen n busness The arcle presens he

More information

2 Aggregate demand in partial equilibrium static framework

2 Aggregate demand in partial equilibrium static framework Unversy of Mnnesoa 8107 Macroeconomc Theory, Sprng 2009, Mn 1 Fabrzo Perr Lecure 1. Aggregaon 1 Inroducon Probably so far n he macro sequence you have deal drecly wh represenave consumers and represenave

More information

A New Method for Computing EM Algorithm Parameters in Speaker Identification Using Gaussian Mixture Models

A New Method for Computing EM Algorithm Parameters in Speaker Identification Using Gaussian Mixture Models 0 IACSI Hong Kong Conferences IPCSI vol. 9 (0) (0) IACSI Press, Sngaore A New ehod for Comung E Algorhm Parameers n Seaker Idenfcaon Usng Gaussan xure odels ohsen Bazyar +, Ahmad Keshavarz, and Khaoon

More information

Chapter 3: Signed-rank charts

Chapter 3: Signed-rank charts Chaer : gned-ran chars.. The hewhar-ye conrol char... Inroducon As menoned n Chaer, samles of fxed sze are aen a regular nervals and he long sasc s hen loed. The queson s: Whch qualy arameer should be

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

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

Learning Objectives. Self Organization Map. Hamming Distance(1/5) Introduction. Hamming Distance(3/5) Hamming Distance(2/5) 15/04/2015

Learning Objectives. Self Organization Map. Hamming Distance(1/5) Introduction. Hamming Distance(3/5) Hamming Distance(2/5) 15/04/2015 /4/ Learnng Objecves Self Organzaon Map Learnng whou Exaples. Inroducon. MAXNET 3. Cluserng 4. Feaure Map. Self-organzng Feaure Map 6. Concluson 38 Inroducon. Learnng whou exaples. Daa are npu o he syse

More information

HEURISTICS FOR ONE CLASS OF MINIMAL COVERING PROBLEM IN CASE OF LOCATING UNDESIRABLE GOODS

HEURISTICS FOR ONE CLASS OF MINIMAL COVERING PROBLEM IN CASE OF LOCATING UNDESIRABLE GOODS HEURSTCS FOR ONE CLASS OF MNMAL COVERNG PROBLEM N CASE OF LOCATNG UNDESRABLE GOODS BRANKA DMTRJEVC MLORAD VDOVC The Faculy of Transpor and Traffc Engneerng, The Unversy of Belgrade, Belgrade, Serba and

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

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

Displacement, Velocity, and Acceleration. (WHERE and WHEN?)

Displacement, Velocity, and Acceleration. (WHERE and WHEN?) Dsplacemen, Velocy, and Acceleraon (WHERE and WHEN?) Mah resources Append A n your book! Symbols and meanng Algebra Geomery (olumes, ec.) Trgonomery Append A Logarhms Remnder You wll do well n hs class

More information

グラフィカルモデルによる推論 確率伝搬法 (2) Kenji Fukumizu The Institute of Statistical Mathematics 計算推論科学概論 II (2010 年度, 後期 )

グラフィカルモデルによる推論 確率伝搬法 (2) Kenji Fukumizu The Institute of Statistical Mathematics 計算推論科学概論 II (2010 年度, 後期 ) グラフィカルモデルによる推論 確率伝搬法 Kenj Fukuzu he Insue of Sascal Maheacs 計算推論科学概論 II 年度 後期 Inference on Hdden Markov Model Inference on Hdden Markov Model Revew: HMM odel : hdden sae fne Inference Coue... for any Naïve

More information

Advanced Macroeconomics II: Exchange economy

Advanced Macroeconomics II: Exchange economy Advanced Macroeconomcs II: Exchange economy Krzyszof Makarsk 1 Smple deermnsc dynamc model. 1.1 Inroducon Inroducon Smple deermnsc dynamc model. Defnons of equlbrum: Arrow-Debreu Sequenal Recursve Equvalence

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

A capacitor consists of two conducting plates, separated by an insulator. Conduction plates: e.g., Aluminum foil Insulator: air, mica, ceramic, etc

A capacitor consists of two conducting plates, separated by an insulator. Conduction plates: e.g., Aluminum foil Insulator: air, mica, ceramic, etc 3//7 haper 6 apacors and Inducors Makng preparaon for dynamc crcus, whch hae far more applcaons han he sac crcus we hae learned so far. 6. apacors Sore energy n elecrc feld nsulaor onducng plaes A capacor

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

3. OVERVIEW OF NUMERICAL METHODS

3. OVERVIEW OF NUMERICAL METHODS 3 OVERVIEW OF NUMERICAL METHODS 3 Inroducory remarks Ths chaper summarzes hose numercal echnques whose knowledge s ndspensable for he undersandng of he dfferen dscree elemen mehods: he Newon-Raphson-mehod,

More information

Endogeneity. Is the term given to the situation when one or more of the regressors in the model are correlated with the error term such that

Endogeneity. Is the term given to the situation when one or more of the regressors in the model are correlated with the error term such that s row Endogeney Is he erm gven o he suaon when one or more of he regressors n he model are correlaed wh he error erm such ha E( u 0 The 3 man causes of endogeney are: Measuremen error n he rgh hand sde

More information

Determining the Number of Games Needed to Guarantee an NHL Playoff Spot

Determining the Number of Games Needed to Guarantee an NHL Playoff Spot Deermnng he Number of Games Needed o Guaranee an NHL Playoff Spo Tyrel Russell and Peer van Bee Cheron School of Compuer Scence Unversy of Waerloo {crussel,vanbee}@uwaerloo.ca Absrac. Many spors fans nves

More information

Modeling and Solving of Multi-Product Inventory Lot-Sizing with Supplier Selection under Quantity Discounts

Modeling and Solving of Multi-Product Inventory Lot-Sizing with Supplier Selection under Quantity Discounts nernaonal ournal of Appled Engneerng Research SSN 0973-4562 Volume 13, Number 10 (2018) pp. 8708-8713 Modelng and Solvng of Mul-Produc nvenory Lo-Szng wh Suppler Selecon under Quany Dscouns Naapa anchanaruangrong

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

EE 247B/ME 218: Introduction to MEMS Design Lecture 27m2: Gyros, Noise & MDS CTN 5/1/14. Copyright 2014 Regents of the University of California

EE 247B/ME 218: Introduction to MEMS Design Lecture 27m2: Gyros, Noise & MDS CTN 5/1/14. Copyright 2014 Regents of the University of California MEMSBase Fork Gyrosoe Ω r z Volage Deermnng Resoluon EE C45: Inrouon o MEMS Desgn LeM 15 C. Nguyen 11/18/08 17 () Curren (+) Curren Eleroe EE C45: Inrouon o MEMS Desgn LeM 15 C. Nguyen 11/18/08 18 [Zaman,

More information

A Tour of Modeling Techniques

A Tour of Modeling Techniques A Tour of Modelng Technques John Hooker Carnege Mellon Unversy EWO Semnar February 8 Slde Oulne Med neger lnear (MILP) modelng Dsuncve modelng Knapsack modelng Consran programmng models Inegraed Models

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

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

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

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

Bayes rule for a classification problem INF Discriminant functions for the normal density. Euclidean distance. Mahalanobis distance

Bayes rule for a classification problem INF Discriminant functions for the normal density. Euclidean distance. Mahalanobis distance INF 43 3.. Repeon Anne Solberg (anne@f.uo.no Bayes rule for a classfcaon problem Suppose we have J, =,...J classes. s he class label for a pxel, and x s he observed feaure vecor. We can use Bayes rule

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

An introduction to Support Vector Machine

An introduction to Support Vector Machine An nroducon o Suppor Vecor Machne 報告者 : 黃立德 References: Smon Haykn, "Neural Neworks: a comprehensve foundaon, second edon, 999, Chaper 2,6 Nello Chrsann, John Shawe-Tayer, An Inroducon o Suppor Vecor Machnes,

More information

, t 1. Transitions - this one was easy, but in general the hardest part is choosing the which variables are state and control variables

, t 1. Transitions - this one was easy, but in general the hardest part is choosing the which variables are state and control variables Opmal Conrol Why Use I - verss calcls of varaons, opmal conrol More generaly More convenen wh consrans (e.g., can p consrans on he dervaves More nsghs no problem (a leas more apparen han hrogh calcls of

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

A. Inventory model. Why are we interested in it? What do we really study in such cases.

A. Inventory model. Why are we interested in it? What do we really study in such cases. Some general yem model.. Inenory model. Why are we nereed n? Wha do we really udy n uch cae. General raegy of machng wo dmlar procee, ay, machng a fa proce wh a low one. We need an nenory or a buffer or

More information

M. Y. Adamu Mathematical Sciences Programme, AbubakarTafawaBalewa University, Bauchi, Nigeria

M. Y. Adamu Mathematical Sciences Programme, AbubakarTafawaBalewa University, Bauchi, Nigeria IOSR Journal of Mahemacs (IOSR-JM e-issn: 78-578, p-issn: 9-765X. Volume 0, Issue 4 Ver. IV (Jul-Aug. 04, PP 40-44 Mulple SolonSoluons for a (+-dmensonalhroa-sasuma shallow waer wave equaon UsngPanlevé-Bӓclund

More information

Dynamic Team Decision Theory

Dynamic Team Decision Theory Dynamc Team Decson Theory EECS 558 Proec Repor Shruvandana Sharma and Davd Shuman December, 005 I. Inroducon Whle he sochasc conrol problem feaures one decson maker acng over me, many complex conrolled

More information

Introduction to Boosting

Introduction to Boosting Inroducon o Boosng Cynha Rudn PACM, Prnceon Unversy Advsors Ingrd Daubeches and Rober Schapre Say you have a daabase of news arcles, +, +, -, -, +, +, -, -, +, +, -, -, +, +, -, + where arcles are labeled

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

Chapter 5. Circuit Theorems

Chapter 5. Circuit Theorems Chaper 5 Crcu Theorems Source Transformaons eplace a olage source and seres ressor by a curren and parallel ressor Fgure 5.-1 (a) A nondeal olage source. (b) A nondeal curren source. (c) Crcu B-conneced

More information