Asymptotic Optimality of Order-up-to Policies in Lost Sales Inventory Systems

Size: px
Start display at page:

Download "Asymptotic Optimality of Order-up-to Policies in Lost Sales Inventory Systems"

Transcription

1 Asympoic Opimaliy of Order-up-o Policies in Los Sales Invenory Sysems Woonghee Tim Huh 1, Ganesh Janakiraman 2, John A. Mucksad 3, Paa Rusmevichienong 4 December 4, 2006 Subjec Classificaion: Invenory/Producion: Los Sales, Backorders, Finie and Infinie Horizon, Opimal Coss Absrac We sudy a single-produc single-locaion invenory sysem under periodic review, where excess demand is los and he replenishmen lead ime is posiive. The performance measure of ineres is he long run average holding cos and los sales penaly. For a large class of demand disribuions, we show ha when he los sales penaly becomes large compared o he holding cos, he relaive difference beween he cos of he opimal policy and he bes order-up-o policy converges o zero. For any given cos parameers, we esablish a bound on his relaive difference. Numerical experimens show ha he bes order-up-o policy performs well, yielding an average cos ha is wihin 1.5% of he opimal cos even when he raio beween he los sales penaly and he holding cos is jus Deparmen of Indusrial Engineering and Operaions Research, Columbia Universiy, New York, NY 10027, USA. huh@ieor.columbia.edu. 2 IOMS-OM Group, Sern School of Business, New York Universiy, 44 W. 4h Sree, Room 8-71, New York, NY gjanakir@sern.nyu.edu 3 School of Operaions Research and Indusrial Engineering, Cornell Universiy, Ihaca, NY 14853, USA. jack@orie.cornell.edu 4 School of Operaions Research and Indusrial Engineering, Cornell Universiy, Ihaca, NY 14853, USA. paarus@cornell.edu 1

2 1. Inroducion We sudy he opimal replenishmen policy in a periodic-review single-sage invenory sysem ha procures invenory from a source wih ample supply. There is a replenishmen lead ime of τ periods beween placing an order and is delivery. Demands in differen periods are independen and idenically disribued. In he even ha demand in a period exceeds he available on-hand invenory, excess demand is los and we incur a los sales penaly cos of $b per uni. We also charge holding coss on invenory on hand a he end of each period a he rae of $h per uni per period. We wish o find an ordering policy ha minimizes he long run average holding cos and los sales penaly. When demand is backordered insead of los, Karlin and Scarf (1958) show ha an order-up-o policy is opimal; ha is, here is an order-up-o-level o which we raise he invenory posiion defined as invenory available for immediae sales plus he amoun of invenory ha has been ordered and no ye delivered in each period. They also show ha his simple policy fails o be opimal in he los sales model. In many business environmens, los sales occur frequenly when cusomer demands are no me immediaely. Hence, finding he opimal replenishmen policy, characerizing is srucural properies, and developing heurisics ha work well in pracical seings are imporan. Moreover, in many imporan invenory sysems, we observe ha he los sales penaly b is generally much larger han he holding cos h, as shown in he following examples from reail and service pars environmens. In his paper, we propose simple invenory policies ha are guaraneed o perform well in such sysems. Reail: Consider a produc whose procuremen cos is $1 per uni o he reailer. Assume he reailer reviews he sysem and replenishes is invenory once a week, and sells he produc a $(1 + m) per uni, where m represens he mark-up. The los sales cos in his case is a leas $m per uni, no including any loss in cusomer goodwill due o unfulfilled demand. The holding cos per uni per period is simply he cos of holding $1 in invenory for a week. A a cos of capial of 15% per year, his is approximaely $ per uni per period. So, he raio beween he los sales penaly cos and he holding cos in his example is a leas 400m. A a 25% mark-up, which is quie common in many reail environmens, his raio is a leas

3 Service Pars Mainenance: Consider, for example, he business of mainaining service pars for personal compuers, phoocopiers, or elecommunicaion equipmens. Mos corporae cliens purchase service-level agreemens ha require he manufacurer, in he even of a failure, o bring he equipmen back o service wihin a specified ime window wihin sipulaed minimum probabiliies (for example, wihin 2 hours for 95% of failures and wihin 6 hours for 99% of failures). To mee hese agreemens, he equipmen manufacurers frequenly expedie service pars o cusomer locaions when he closes socking locaions do no have he necessary pars. Consider a $100 par ha has o be expedied a an addiional cos of $14. These sysems are ypically reviewed once a day. Assuming a cos of capial of 25% per year, he cos of holding his par in invenory for one day is abou $0.07. Here, he raio beween he los sales penaly cos in his case, he expediing premium of $14 and he holding cos is 200. Our main resul is ha, under mild assumpions on he demand disribuion, he class of orderup-o policies is asympoically opimal for hese sysems as he los sales penaly increases. In fac, we show asympoic opimaliy for a specific order-up-o policy ha is compued using he newsvendor formula wih appropriae parameers. For any given cos parameers, we also esablish an upper bound on he increase in he oal cos from using his specific order-up-o policy insead of he opimal policy. Finally, we presen several compuaional resuls o evaluae he performance of he opimal policy, he bes order-up-o policy, and he specific order-up-o policy menioned above, for a wide range of demand disribuions and cos parameers. 1.1 Noaion and Problem Descripion To faciliae he discussion of our main resuls, le us inroduce he noaion and he problem descripion. We will consider boh he los sales and he backorder sysems. In boh sysems, he lead ime beween he placemen of a replenishmen order and is delivery is denoed by τ. The index for ime periods is and D is he demand in period. We assume D 1, D 2,... are independen and idenically disribued random variables and we use D o denoe a generic random variable wih he same disribuion as D. Also, le D = τ+1 =1 D denoe he oal demand over τ + 1 periods, represening he oal demand over he lead ime including he period when we place he order. Le F denoe he disribuion funcion of D. 3

4 X L A he beginning of period, he replenishmen order placed in period τ is received. Le [0, ) denoe he invenory on hand a his insan in he los sales sysem. For he backorder sysem, le X B (, ) denoe he ne-invenory in period, ha is, he invenory on hand minus backorders a he insan afer receiving he delivery due in period. Afer receiving deliveries, a new replenishmen order is placed afer which he demand D is observed. For any h 0 and b 0, we denoe by L(h, b) he los sales sysem and by B(h, b) he backorder sysem, which is idenical o L(h, b) excep ha excess demand is backordered. In boh he los sales L(h, b) and backorder B(h, b) sysems, we charge holding coss on invenory on hand a he end of each period a he rae of $h per uni per period. While we incur los sales penaly of $b per uni of unme demand in he los sales model L(h, b), he shorage coss in he backorder sysems B(h, b) are charged a he rae of $b per uni of backordered demand per period. Given he holding cos h and los sales penaly b, we denoe by C L,S (h, b) and C L (h, b) he long run average cos in he los sales sysem L(h, b) under an order-up-o-s policy and under an opimal policy, respecively. The corresponding quaniies C B,S (h, b), and C B (h, b) are defined similarly, wih he inerpreaion of b as he backorder cos per uni per period. We denoe by S L (h, b) and S B (h, b) he bes order-up-o levels in he los sales L(h, b) and he backorder B(h, b) sysems, respecively. We noe ha in he backorder sysem B(h, b), order-up-o policies are opimal, and he bes order-up-o level is given by he newsvendor formula under he disribuion funcion F of D. 1.2 Conribuions and Organizaion of he Paper Our analysis provides some imporan insighs abou boh los sales and backorder invenory sysems, in addiion o he main resul on he asympoic opimaliy of order-up-o policies in los sales sysems. We now describe he organizaion of he paper and discuss he main conribuions of he individual secions. Table 1 provides a summary of he main resuls. In Secion 2, we provide a brief lieraure review and indicae how our research conribues o he curren research on los sales sysems. In Secion 3, we describe he assumpion on he demand disribuion ha we will use hroughou he paper. We hen show in Theorem 4 ha our assumpion encompasses a broad class of demand disribuions ha commonly occur in many invenory seings. 4

5 Caegory Descripion Resuls Backorder Robusness of Opimal For any ν > 0, lim b C B (h,νb) C B (h,b) Sysem Cos and Newsvendor lim b C B,S νb(h,b) C B (h,b) = 1, and = lim b C B,S b(h,νb) C B (h,νb) = 1, (Secion 4) Soluion (Theorem 6) where S b = S B (h, b) and S νb = S B (h, νb) Asympoic Equivalence Connecions of he Opimal Coss lim b C B (h,b) C L (h,b) = 1 Beween (Theorem 8) Los Sales Bounds on he Cos For any S, and of Any Order-up-o C B,S (h, b/(τ + 1)) C L,S (h, b) C B,S (h, b + τh) Backorder Policy (Lemma 13) Sysems Bounds on he Bes (Secion 5) Order-up-o Levels S B (2h(τ + 1), b h(τ + 1)) S L (h, b) S B (h, b + τh) (Theorem 14) Asympoic Opimaliy Main Resuls of Order-up-o Policies lim b min S C L,S (h,b) C L (h,b) (Secion 6) in Los Sales Sysems where S b+τh = S B (h, b + τh). (Theorem 15) = lim b C L,S b+τh(h,b) C L (h,b) = 1, Table 1: A summary of he main resuls in he paper. All asympoic resuls assume ha he disribuion of he demand over lead ime saisfies Assumpion 1, which is discussed in deail in Secion 3. Robusness of he Opimal Cos and he Opimal Policy in Backorder Sysems (Secion 4): As our firs conribuion, we show ha he opimal cos in he backorder sysem is robus agains changes in he backorder cos parameer b for large b. More precisely, he increase in oal cos resuling from incorrecly esimaing b becomes negligible for large b; ha is, for any h 0 and ν > 0, C B,S νb(h, b) C B (h, νb) lim b C B = lim (h, b) b C B (h, b) = lim C B,S b(h, νb) b C B (h, νb) = 1, where S b = S B (h, b) and S νb = S B (h, νb) denoe he opimal order-up-o levels in he backorder sysems B(h, b) and B(h, νb), respecively. Esimaing he backorder cos can be difficul in many applicaions because we have o assess he long-erm impac of a sockou and accoun for losses in cusomer goodwill from delays in order 5

6 fulfillmen. Suppose we misakenly esimae he backorder parameer o be νb (insead of b) and use he order-up-o-s νb policy in he B(h, b) sysems. The above resul shows ha he relaive increase in he oal cos from using such a policy converges o zero as b increases. We hus make precise he folk heorem ha he cos funcion in a ypical invenory conrol problem is fla around he opimal soluion. Ineresingly, he above resul holds only for demand disribuions saisfying Assumpion 1 (see Secion 3 for more deails). In Secion 4.1, we provide a counerexample ha does no saisfy Assumpion 1 and where he above resul fails. Connecions Beween Los Sales and Backorder Sysems (Secion 5): As our second conribuion, we also esablish relaionships beween los sales and backorder sysems, in erms of he opimal cos, he cos of any order-up-o policy, and he bes order-up-o level. In Theorem 8 in Secion 5, we show he asympoic equivalence beween he opimal cos in he los sales and he backorder sysems as he parameer b increases; ha is, for any h 0, C B (h, b) lim b C L (h, b) = 1. When he parameer b is large, his resul enables us o use he (easily compued) opimal cos of he backorder sysem B(h, b) as an approximaion for he opimal cos in he corresponding los sales sysem. In addiion o asympoic equivalence of he opimal coss, he long run average cos of any order-up-o policy in he los sales L(h, b) sysem is bounded above and below by he cos of he same policy in he backorder sysems B(h, b + τh) and B(h, b/(τ + 1)), respecively. Lemma 13 shows ha for any order-up-o level S, C B,S (h, b/(τ + 1)) C L,S (h, b) C B,S (h, b + τh). We also develop bounds on he bes order-up-o level in he los sales sysem, as shown in Theorem 14, ha S B (2h(τ + 1), b h(τ + 1)) S L (h, b) S B (h, b + τh). The above bounds represen he firs such resuls ha relae he cos of any order-up-o policy and he bes order-up-o level in he los sales sysem wih he corresponding quaniies in he wellsudied backorder sysem. These bounds are easily compuable since hey are simply represened by newsvendor formulas. Main Resuls (Secion 6): The resuls from Secion 4 and 5 se he sage for he main resul of he paper (Theorem 15): order-up-o policies are asympoically opimal in he los sales sysem, 6

7 or min S C L,S (h, b) C L,Sb+τh(h, b) lim b C L = lim (h, b) b C L = 1, (h, b) where S b+τh = S B (h, b + τh). The above resul shows ha, in fac, he opimal order-up-o level for he backorder sysem B(h, b + τh) is asympoically opimal for he los sales sysem L(h, b). Theorem 15 also provides an explici and compuable bound on he rae of convergence for any finie value of b. Compuaional Invesigaion (Secion 7): In addiion o esablishing asympoic opimaliy of base sock policies in los sales invenory sysems, we also discuss exensive compuaional invesigaion. In our experimen, we indicae he performance of base sock policies under differen problem parameers. We show how order-up-o policies perform agains oher replenishmen heurisics (Secion 7.2), deermine he impac of increasing demand on oal cos (Secion 7.3), and sudy how well order-up-o policies would do when he demand exhibis high variance-o-mean raios (Secion 7.4). Our compuaional resuls show ha he cos of he bes order-up-o policy is wihin 1.5% of he opimal cos even when he raio beween he los sales penaly and he holding cos is jus 100. Moreover, our order-up-o policy coninues o perform well even for demands wih large means or high variance. This resul suggess ha such a policy should perform well in many pracical invenory applicaions. 2. Brief Lieraure Review There are hree main sreams of research on los sales invenory sysems: he analysis of he opimal invenory policy, he analysis of hese sysems under an arbirary policy or under policies of specific kinds, and he compuaional invesigaion of he performance of easily implemenable heurisics. We now briefly review hese hree research sreams in ha order. Karlin and Scarf (1958) firs sudy he los sales invenory sysem wih a lead ime of one period. They demonsrae ha order-up-o policies are no opimal for hese sysems; he opimal order quaniy is a decreasing funcion of he amoun of invenory on hand, wih he rae of decrease being smaller han one. For he general lead ime case, hey analyze he sysem under order-upo policies and exponenially disribued demands, and derive an expression for he seady sae disribuion of on-hand invenory level. Moron (1969) exends Karlin and Scarf s resuls on he opimal ordering policy o he general lead ime case. He also derives upper and lower bounds on he opimal order quaniy in a period as funcions of he sae vecor. Recenly, Zipkin (2006b) 7

8 presens an elegan derivaion of Moron s srucural resuls and exends he resuls o more general los sales invenory sysems (for example, allowing capaciy resricions). Levi e al. (2005) develop a heurisic based on he dual balancing echnique inroduced originally for backorder sysems by Levi e al. (2004). They show ha his heurisic aains an expeced cos per period ha is a mos wice ha achieved by an opimal policy for a large class of demand models. Janakiraman e al. (2005) show analyically ha he opimal cos of managing a los sales invenory sysem is smaller han ha of managing a backorder sysem when he backorder cos parameer in he laer sysem has he same magniude as he los sales cos parameer in he former sysem. Under varying assumpions, Karush (1957), Downs e al. (2001) and Janakiraman and Roundy (2004) all consider los sales invenory sysems under order-up-o policies and show he convexiy of he expeced cos per period wih respec o he order-up-o level. Reiman (2004) sudies he class of order-up-o policies and he class of consan order policies (policies ha order a consan quaniy every period regardless of he sae of he sysem). He derives expressions for he order-up-o level and for he consan order-quaniy ha are asympoically opimal wihin he respecive classes of policies as he penaly cos becomes large. He also invesigaes he comparaive performance of he wo policies as he lead ime grows. Moron (1971) compuaionally invesigaes he performance of he myopic policy as a heurisic for problems wih a lead ime of one or wo periods. For los sales problems wih addiional feaures (for example, a se-up cos), Nahmias (1979) derives an inuiively appealing heurisic and invesigaes is performance for he cases of one and wo period lead imes. Recenly, for problems wih lead imes ranging from one o four periods, Zipkin (2006a) invesigaes he performance of he opimal order-up-o policies, he myopic policy, a modified myopic policy ha is based on he coss incurred in wo periods, he dual balancing policy, a generalizaion of base-sock policies suggesed by Moron (1971), and he opimal consan-order policy. To reduce he compuaional effor involved in evaluaing each of hese policies, he presens elegan analyical bounds on he size of he effecive sae space under any given policy. Our paper has elemens of all hree research sreams. Our asympoic opimaliy resuls conribue o an undersanding of he srucure of he opimal policy by esablishing condiions under which he opimal cos is asympoically equal o he cos of he bes base sock policy. Our bounds on he performance of a specific order-up-o policy and he analysis leading o such bounds illuminae he srucural properies of base sock policies and esablish connecions beween los sales 8

9 and backorder invenory sysems. Finally, we complemen Zipkin (2006a) by invesigaing he compuaional performance of order-up-o policies over a larger class of problem insances, especially when he los sales penaly is significanly higher han he holding cos. We show ha when he raio b/h is large, order-up-o policies perform exremely well, wih an average cos ha is wihin 1.5% of he opimal. 3. Assumpion on he Demand Disribuion Recall ha D τ+1 =1 D denoes he oal demand over τ + 1 periods, represening he oal demand over he lead ime including he period when we place he order. Le F denoe he disribuion funcion of D. For any 0, we define he mean residual life m D () as follows: E [ D D > ], if < sup {x : F (x) < 1}, m D () = 0, oherwise. Through ou his paper, we make he following assumpion on he disribuion of D. Assumpion 1. lim m D ()/ = 0. The above assumpion implies ha he expeced mean residual life of D a does no grow faser han. Before proceeding wih examples of demand disribuions saisfying Assumpion 1, le us recall he following definiion. Definiion 2. A coninuous (resp. discree) random variable Y wih a disribuion funcion F and a densiy funcion f (resp. probabiliy mass funcion p) has an increasing failure rae (IFR) propery if f(x)/(1 F (x)) (resp. p(x)/(1 F (x))) is nondecreasing in x. The following resul provides an equivalen characerizaion of an IF R random variable. The proof appears in Secion 1.B.1 of Shaked and Shanhikumar (1994). Lemma 3. A random variable Y is IFR if and only if for any 0 1 < 2, he residual life of Y a 2 is sochasically smaller han he residual life of Y a 1 ; ha is, for any s 0, P { Y 2 > s } { Y > 2 P Y 1 > s } Y > 1. The following heorem, whose proof appears in Appendix A, shows ha Assumpion 1 encompasses a large class of discree and coninuous demand disribuions ha commonly occur in many invenory sysems. 9

10 Theorem 4. If any of he following condiions holds, hen D saisfies Assumpion 1. (a) The demand D in each period (eiher discree or coninuous) is bounded; ha is, here exiss M such ha P {D M} = 1. (b) The demand D in each period (eiher discree or coninuous) has an increasing failure rae (IFR) disribuion. (c) D has a finie variance and he disribuion F of D has a densiy funcion f and a failure rae funcion r() of F ha does no decrease o zero faser han 1/; ha is, where for any 0, r() = f()/(1 F ()). lim r() =, The above heorem shows ha Assumpion 1 encompasses a very large of demand disribuions used in many supply chain models, including any bounded demand random variables. For unbounded demand, par (b) of Theorem 4 shows ha many commonly used disribuions also saisfy Assumpion 1. Examples include geomeric disribuions, Poisson disribuions (see Corollary 5.2 in Ross e al. (2005)), negaive binomial disribuions wih parameer r > 0 and 0 < p < 1, exponenial disribuions, and Gaussian disribuions. When he demand disribuion does no exhibi an IFR propery, par (c) of he above heorem shows ha Assumpion 1 remains saisfied as long as he failure rae does no decrease o zero oo quickly. The following example shows a disribuion ha is no IFR, ye sill saisfies par (c) of Theorem 4. Example 1. Suppose ha D follows a Weibull disribuion wih scale parameer λ > 0 and shape parameer 0 < k < 1. Then, D has he following disribuion, densiy, and failure rae funcions: for any x > 0, f(x) = kxk 1 e (x/λ)k λ k, F (x) = 1 e (x/λ)k, and r(x) = k ( x ) k 1, λ λ and he firs wo momens of D are ( E [D] = λγ ) k and E [ D 2] ( = λ 2 Γ ), k where Γ( ) denoes he Gamma funcion. Since 0 < k < 1, i is easy o verify ha D is no IFR, bu he failure rae funcion r sill saisfies par (c) of Theorem 4. 10

11 4. Asympoic Properies for Backordered Sysems We sar our analysis by showing asympoic properies of he opimal policy in he backorder sysem as he backordere cos parameer b becomes large. To faciliae our discussion, le us inroduce he following noaion. For any y 0, le ψ(y; h, b) denoe he raio beween he expeced backorder and holding coss given he invenory posiion y in he B(h, b) sysem, ha is, ψ(y; h, b) = be [ (D y) +] he [ (y D) +]. The following lemma esablishes upper and lower bounds on he increase in he oal cos from using sub-opimal policies. Lemma 5. For any h 0, b 0, and ν > 0, le S b = S B (h, b) and S νb = S B (h, νb) denoe he opimal policies in he B(h, b) and B(h, νb) backorder sysems, respecively. Then, he relaive difference beween he opimal coss of B(h, b) and B(h, νb) backorder sysems can be bounded as follows: 1 + ψ (S νb ; h, νb) 1 + (1/ν)ψ (S νb ; h, νb) = CB (h, νb) C B,S CB (h, νb) νb (h, b) C B (h, b) CB,Sb(h, νb) C B = 1 + νψ (S b; h, b) (h, b) 1 + ψ (S b ; h, b) Proof. The firs and second inequaliies follows from he fac ha C B (h, b) C B,S νb(h, b) and C B (h, νb) C B,S b(h, νb), respecively. To esablish he firs equaliy, noe ha C B (h, νb) C B,S = νbe [D S νb] + + he [S νb D] + νb (h, b) be [D S νb ] + + he [S νb D] + = 1 + ψ (S νb; h, νb) 1 + (1/ν)ψ (S νb ; h, νb), where he las inequaliy follows from dividing he numeraor and denominaor by he [S νb D] +. The proof of he second equaliy of he lemma is similar. The bounds in Lemma 5 lead direcly o he main asympoic resul of his secion, which is saed in he following heorem. Theorem 6. Under Assumpion 1, he following resuls hold for any h 0. (a) The raio beween he expeced backorder cos per period o he expeced holding cos per period under he opimal policy converges o zero as he backordere cos b increases, ha is, lim ψ ( S B (h, b); h, b ) = 0. b 11

12 (b) For large values of b, he opimal cos and he opimal policy are robus agains changes in he backorder cos; ha is, for any ν > 0, C B,S νb(h, b) C B (h, νb) lim b C B = lim (h, b) b C B (h, b) = lim C B,S b(h, νb) b C B (h, νb) = 1, where S b = S B (h, b) and S νb = S B (h, νb). Proof. To esablish he resul in par (a), noe ha he opimal order-up-o level S B (h, b) is given by he newsvendor formula: S B (h, b) = inf { y : P {D y} b }, b + h which implies ha P { D S B (h, b) } b/(b + h) and P { D > S B (h, b) } h/(b + h) by he lef coninuiy of he disribuion funcion. Therefore, ψ ( S B (h, b); h, b ) = b P { D > S B (h, b) } E [ D S B (h, b) D > S B (h, b) ] h P {D S B (h, b)} E [ S B (h, b) D D S B (h, b) ] E [ D S B (h, b) D > S B (h, b) ] E [ S B (h, b) D D S B (h, b) ] = ( ( md S B (h, b) ) ) ( S B (h, b) S B (h, b) S B (h, b) E [ D D S B (h, b) ] Since E [ D D S B (h, b) ] E [D], he desired resul follows from Assumpion 1 and he definiion of S B (h, b). To prove par (b), noe ha i follows from par (a) and he bounds in Lemma 5 ha lim b C B (h, νb) C B (h, νb) C B,S = lim νb (h, b) b C B (h, b) = lim C B,S b(h, νb) b C B = 1, (h, b) and he desired resul follows from he fac ha ). C B,S νb(h, b) C B (h, b) = CB (h, νb)/c B (h, b) C B (h, νb)/c B,S νb (h, b) and CB,Sb(h, νb) C B (h, νb) = CB,Sb(h, νb)/c B (h, b) C B (h, νb)/c B (h, b). Theorem 6 shows ha, for disribuions saisfying Assumpion 1, he newsvendor soluion and he opimal cos are robus agains inaccurae esimaion of he backorder parameer when he backorder parameer b is large. However, when Assumpion 1 fails, he resul of Theorem 6 many no longer hold, as we now demonsrae. 12

13 4.1 Pareo Disribuions: An Example Where Theorem 6 Fails For any θ > 1, le he densiy funcion f θ be defined by: for any x 0, f θ (x) = θ (1 + x) 1+θ, and le F θ (x) = x 0 f θ(u)du denoe he corresponding disribuion funcion. The following proposiion shows ha F θ does no saisfy Assumpion 1 and he opimal cos is sensiive o he backorder parameer, even for large values of b. Proposiion 7. For any θ > 1, b 0, h 0, and ν > 0, if D has a disribuion funcion F θ, hen Proof. Please see Appendix B. m D () lim = 1 θ 1 and C B (h, νb) lim b C B (h, b) = ν1/θ. 5. Connecions beween Los Sales and Backorder Invenory Sysems The following resul provides an inuiive basis for conjecuring ha he opimal policy in he backorder sysem is also asympoically opimal in he los sales sysem. Theorem 8. Under Assumpion 1, as b increases, he opimal cos in he los sales L(h, b) sysem converges o he opimal cos in he backorder sysem B(h, b); ha is, for any h 0, C L (h, b) lim b C B (h, b) = 1. Proof. Janakiraman e al. (2005) esablished he following bounds on he opimal cos in he los sales sysem: C B (h, b/(τ + 1)) C L (h, b) C B (h, b). From Theorem 6(b), lim b C B (h, b)/c B (h, b/(τ + 1)) = 1, which gives he desired resul. Nex, we esablish connecions beween he dynamics in boh he los sales and he backorder sysems under he same order-up-o policy. Le X L,S and X B,S denoe he on-hand invenory in he los sales sysem and he ne invenory in he backorder sysem, respecively, a he beginning of period under an order-up-o-s policy. Similarly, we use and BACK B,S o denoe he los sales incurred in period and he backorders exised a he end of period, respecively, under he order-up-o-s policy. By definiion, X L,S, and BACK B,S are non-negaive random variables. The following lemma esablishes he relaionship among hese random variables. 13

14 Lemma 9. Assume boh he los sales and he backorder sysems sar a he same sae in period 1 and he invenory posiion in his sae is S or less. Then, for every demand sample pah and for every τ + 1, X B,S X L,S and BACK B,S 1 i= τ i BACK B,S. Proof. I follows from he dynamics of los sales sysems under he order-up-o-s policy (see Janakiraman and Roundy (2004)) ha X L,S = S 1 i= τ D i + 1 i= τ i = X B,S + 1 i= τ i, for any τ + 1, which proves he firs par of he lemma. Since x + y (x y) + for any x R and y 0, we have BACK B,S 1 i= τ i = = ( ( D X B,S D X L,S ) + 1 i= τ ) + = LOST L,S where he las inequaliy follows from he fac ha X B,S i ( X L,S. ( D X B,S D X B,S 1 i= τ ) + = BACK B,S, i ) + The resul of Lemma 9 relaes he ne invenory in he backorder sysem wih he on-hand invenory in he los sales sysem for any finie ime period. To use his resul for sudying he long run average cos in L(h, b), his resul should be exended o he seady-sae on-hand invenory, which we will denoe by X L,S. This is our nex sep. Before proving properies of X L,S well known (and rivial o verify) ha X B,S and X B,S, i is imporan o esablish heir exisence. I is exiss and τ D. X B,S d S However, in general, for any given saring sae and order-up-o level S, i is no rue ha he disribuion of X L,S =1 converges o a saionary disribuion. In fac, Huh e al. (2006) give such an example. Ineresingly, we are able o show wo resuls ha help us resolve his difficuly. Lemma 10. For every S and any saring sae in period 1, he sequence of he expeced cos per period over he inerval [1, T ] given by T =1 E[h (XL,S D ) + + b (D X L,S ) + ] T converges o a limi ha is independen of he saring sae. 14

15 Proof. Please see Appendix C. Since he long run average cos is he quaniy of ineres o us in his paper, Lemma 10 implies ha we can limi our analysis o any specific saring sae. In he nex lemma, we show ha for a specific saring sae we choose, he saionary disribuion of he on hand invenory, {X L,S }, exiss. Lemma 11. Assume he saring sae (in period 1) is such ha here are S/(τ + 1) unis on hand and S/(τ + 1) unis due o be delivered in each of he periods 2,..., τ. Then, he sequence of he disribuions of he random variables {X L,S } converges. Proof. Please see Appendix D. We will use X L,S o denoe a random variable whose disribuion is he limiing disribuion from Lemma 11. We can now define C L,S (h, b) mahemaically as follows: [ (X C L,S L,S (h, b) = he D ) ] + + be [ (D ) X L,S + ], where he random variable D denoes he demand in a single period. Corollary 12. The random variable X B,S and he random variable i.e. for any z 0, P { X B,S is sochasically smaller han he random variable X L,S is sochasically smaller han he random variable BACK B,S, > z } P { X L,S > z } and P { LOST L,S > z } P { BACK B,S > z }. Proof. Noice ha we assumed ha X L,S represens he limiing disribuion of X L,S when he saring sae vecor has S/(τ + 1) unis in each componen. So, his saring sae saisfies he assumpion of Lemma 9. The resul follows direcly from his lemma. The nex resul esablishes upper and lower bounds on he cos of any order-up-o policy in he los sales sysem in erms of he coss in he backorder sysem. Lemma 13. The long run average cos of he order-up-o-s policy in he los sales sysem L(h, b) is bounded above (resp. below) by he cos of he same policy in he backorder sysem B(h, b + τh) (resp. B(h, b/(τ + 1))), i.e. C B,S (h, b/(τ + 1)) C L,S (h, b) C B,S (h, b + τh). 15

16 Proof. Recall ha he random variable D denoes he oal demand over τ + 1 periods and X L,S = S 1 i= τ D i + 1 i= τ [ ( ) ] + he X L,S D + be [ = he X L,S = he = he [ [ D ] + he S S ( = he S 1 i= τ i= τ i= τ D i + L,S LOSTi. Then, for any, we have [ ( ) ] + D X L,S [ ( ) ] + D X L,S + be ] 1 i= τ D i ] + h = he [ (S D) +] [ he Since he sochasic process i i= τ [ E D i i= τ [ ( ) ] + D X L,S [ + he ] [ + be ) + ( ) + D i he D i S + h BACK B,S ] + h { } X L,S : 1 i= τ [ E C L,S (h, b) = he [ (S D) +] he [ BACK B,S i ] [ + be ] i= τ [ E ] [ + be converges o X L,S, we have i ] ] [ + be ]. ] [ ] + (b + (τ + 1)h)E LOST L,S he [ (S D) +] + (b + τh)e [ BACK B,S ] = C B,S (h, b + τh), [ where he inequaliy follows from Corollary 12 which implies ha E This esablishes he upper bound in he saemen of he lemma. Nex, we esablish he lower bound. Since BACK B,S 1 L,S i= τ LOSTi ] [ E BACK B,S probabiliy one by Lemma 9, aking he expecaion on boh sides and aking he limi as increases [ ] [ ] / o infiniy, i follows ha E E (τ + 1). Therefore, i follows from he above expression for C L,S (h, b) ha BACK B,S C L,S (h, b) he [ (S D) +] ( ) b + (τ + 1)h + h E [ BACK B,S ] = C B,S (h, b/(τ + 1)). τ + 1 ] ]. wih We now relae he bes order-up-o level in he los sales and backorder sysems. Theorem 14. For any h 0 and b 0, he bes order-up-o level in he los sales sysem L(h, b) is 16

17 (a) bounded above by he bes order-up-o level in he backorder sysem B(h, b + τh) wih a backorder penaly cos parameer of b + τh, ha is, S L (h, b) S B (h, b + τh) ; (b) bounded below by he bes order-up-o level in he backorder sysem B(2h(τ + 1), b h(τ + 1)) wih a holding cos parameer of 2h(τ+1) and a backorder penaly cos parameer of b h(τ+1), ha is, S L (h, b) S B (2h(τ + 1), b h(τ + 1)). Proof. Please see Appendix E and Appendix F. 6. Performance Bounds and Asympoic Resuls for Los Sales Sysems We will now esablish he asympoic opimaliy of order-up-o policies in he los sales sysem. We will, in fac, show he asympoic opimaliy of he order-up-o-s B (h, b+τh) policy, corresponding o he opimal policy in he backorder sysem B(h, b + τh). Noice ha his is he upper bound we derived for he bes order-up-o level for he los sales sysem (Theorem 14(a)). For any finie b, we also derive an upper bound on he loss in performance from using his policy relaive o he opimal policy. Recall ha for any y 0, ψ(y; h, b) = be [ (D y) +] he [ (y D) +]. The main resul of his secion is saed in he following heorem. Theorem 15. For any h 0 and b 0, le S b+τh = S B (h, b+τh) and S b/(τ+1) = S B (h, b/(τ +1)) denoe he opimal order-up-o policies in he backorder sysems B(h, b + τ h) and B(h, b/(τ + 1)), respecively. Then, (a) The raio beween he cos of he order-up-o-s b+τh and he opimal policies in he los sales sysem L(h, b) can be bounded as follows: C L,S b+τh(h, b) C L (h, b) 1 + ( (b+τh)(τ+1) b ) ψ ( S b/(τ+1) ; h, b/(τ + 1) ) 1 + ψ ( S b/(τ+1) ; h, b/(τ + 1) ). (b) Under Assumpion 1, he order-up-o-s b+τh policy is asympoically opimal in he los sales sysem L(h, b), i.e. min S 0 C L,S (h, b) lim b C L (h, b) = lim b C L,Sb+τh(h, b) C L (h, b) = 1. 17

18 Proof. We know from Janakiraman e al. (2005) ha C B (h, b/(τ + 1)) C L (h, b) and i follows from Lemma 13 ha C L,S b+τh(h, b) C L (h, b) where ν b = (b + τh)(τ + 1)/b. CB,S b+τh(h, b + τh) C B (h, b/(τ + 1)) = CB (h, b + τh) C B (h, b/(τ + 1)) 1 + ν ( bψ Sb/(τ+1) ; h, b/(τ + 1) ) 1 + ψ ( S b/(τ+1) ; h, b/(τ + 1) ), Noe ha he equaliy follows from he definiion of S b+τh and he las inequaliy follows from Lemma 5. This proves par (a). Since lim b ν b = τ + 1 and by Theorem 6(a) i follows ha which is he desired resul. C L,S b+τh(h, b). lim ψ ( S b/(τ+1) ; h, b/(τ + 1) ) = 0, b C L,S b+τh(h, b) lim b C L = 1, (h, b) To complee he proof, noe ha C L (h, b) min S C L,S (h, b) 7. Compuaional Invesigaion In his secion, we compare he oal cos of he opimal policy, he bes base sock policy, and he base sock policy suggesed in Theorem 15. In Secion 7.1, we describe he mehodologies for our experimens. In Secion 7.2, we consider he performance of our proposed order-up-o policies on problem insances considered by Zipkin (2006b), enabling us o benchmark our performance agains oher replenishmen policies. Then, in Secion 7.3, we sudy he performance of base sock policies as he expeced demand increases, focusing on he commonly used Poisson demand models. Finally, in Secion 7.4, by considering negaive binomial disribuions, we explore he impac of increasing variance-o-mean raios on he performance of base sock policies. 7.1 Mehodologies To compue he long run average cos of he opimal replenishmen policy for a given discree demand disribuion, we consider he average cos dynamic programming formulaion. The sae space S for our dynamic program consiss of τ-dimensional vecors given by S = {(z 0, z 1,..., z τ 1 ) : z i Z + {0}}, where z 0 denoes he on-hand invenory afer receiving he replenishmen order and for 1 i τ 1, z i denoes he replenishmen quaniies ha will arrive i periods from now, corresponding o he 18

19 order placed τ i periods in he pas. In our experimen, he demand D in each period has he propery ha P {D = 0} > 0. I follows ha from Proposiion 2.6 in Bersekas (1995) ha for any h 0 and b 0, he opimal cos C L (h, b) is he unique soluion of he following average cos dynamic program: for any x S, C L { (h, b) + g(x) = min he [z0 D] + + be [D z 0 ] + + E [ g ( (z 0 D) + + z 1, z 2,..., z τ 1, u )]}, u 0 where g( ) denoes he differenial cos vecor and u represens he ordering quaniy. Since we will consider unbounded demand in our experimen, we apply he elegan sae-space reducion echnique inroduced by Zipkin (2006b), enabling us o consider a dynamic program wih only a finie number of saes (alhough he size of he sae space sill increases exponenially wih he lead ime). Then, o deermine he opimal replenishmen policy, we hen apply he relaive value ieraion mehod for 1000 ieraions or unil he change beween ieraions is less han (see Bersekas (1995); Zipkin (2006b) for more deails). We noe ha for any S 0, a similar dynamic programming formulaion can be used o deermine he long run average cos of he order-up-o-s policy. In his case, for any x S, insead of minimizing over all possible ordering quaniies as in opimal dynamic program above, [ he ordering quaniy is given by u = S τ 1 +. i=0 i] z We can hen apply he same relaive value ieraion mehod o deermine he long run average cos C L,S (h, b). To deermine he bes order-up-o level, we use he fac ha he oal cos C L,S (h, b) is convex in S (Downs e al. (2001) and Janakiraman and Roundy (2004)) and he bes order-up-o level is bounded above by S B (h, b + τh) (Theorem 14(a)) and below by S B (2h(τ + 1), b h(τ + 1)) (Theorem 14(b)). 7.2 Represenaive Problems In his secion, we repor he compuaional resuls for represenaive problems considered in Zipkin (2006b), enabling us o compare he cos of our base sock policies wih oher replenishmen heurisics. We consider Poisson and Geomeric demand disribuions, boh wih mean 5. The lead ime ranges from 1 o 4 periods. Assuming a holding cos of $1, we consider he los sales penaly ranging from $1 o $199. We compare he cos of he opimal policy, he bes base sock policy, and he order-up-o-s B (h, b + τh) policy, which is shown o be asympoically opimal (Theorem 15). Table 2 and 3 show he coss of hese polices for Poisson and geomeric disribuions, respecively. (We remark ha Zipkin (2006b) repors he cases where he los sales penaly is $4, $9 or $19.) 19

20 From Table 2 and 3, we observe ha as he los sales penaly increases, he cos of he bes base sock and he order-up-o-s B (h, b + τh) policies converge o he opimal cos as prediced by Theorem 15. For b = 199, he coss of boh base sock policies differ from he opimal by a mos 5%. However, for a specific cos parameer, he performance of our base sock policies ends o degrade as he lead ime increases. When comparing wih he performance of oher heurisics on he same problem insances (as repored in Zipkin (2006b)), he performance of our base sock policies are comparable wih oher heurisics. 7.3 Impac of Varying Mean Demand In his secion, we explore he performance of base sock policies as he mean demand changes. To faciliae our discussion, we assume a lead ime of wo periods (τ = 2) and consider a Poisson demand disribuion whose mean varies from 1 o 10. Table 4 shows a comparison among he opimal cos, he cos of he bes base sock policy, and he order-up-o-s B (h, b + τh) policy. From Table 4, we observe ha, for a specific los sales penaly, he relaive difference beween he opimal cos and he cos of he order-up-o-s B (h, b + τh) policy remains prey small even when he mean demand increases. We observe a similar paern for differen values of lead imes as well. This observaion has an imporan pracical implicaion. When he mean demand is small, compuing he bes order-up-o level is relaively easy since he range of base sock levels o consider is small; in fac, he opimal policy iself migh be compuaionally feasible. On he oher hand, for large means, compuing he opimal policy (or even he bes orderup-o level) is compuaionally more difficul because he search space is larger. Noneheless, our experimenal resuls indicae ha he simple and easily compuable base sock policy (wih he order-up-o level of S B (h, b + τh)) coninues o perform well even his seing, yielding oal cos ha is wihin 2% of he opimal (for b = 199). This resul suggess a pracical and effecive replenishmen heurisic: compue he opimal order-up-o policy exacly for small demand, and use he base sock level S B (h, b + τh) as an approximaion for larger demand. 7.4 Impac of Increasing Variance-o-Mean Raio In his secion, we explore he impac of he variance-o-mean raio on he performance of base sock policies. As in he previous secion, we assume he lead ime is 2 (τ = 2) and he demand D in each period follows a negaive binomial demand disribuion wih parameer (r, p) where r {1, 2} and 20

21 Lead Los Opimal Bes Base Sock S B (h, b + τh) Time Sales Cos Level Cos % Diff From Level Cos % Diff From Penaly Opimal Cos Opimal Cos % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % Table 2: Performance of base sock policies for he Poisson disribuion wih mean 5. 21

22 Lead Los Opimal Bes Base Sock S B (h, b + τh) Time Sales Cos Level Cos % Diff From Level Cos % Diff From Penaly Opimal Cos Opimal Cos % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % Table 3: Performance of base sock policies for he Geomeric disribuion wih mean 5. 22

23 Los Mean Opimal Bes Base Sock S B (h, b + τh) Sales Demand Cos Level Cos % diff from Level Cos % diff from Penaly opimal opimal % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % Table 4: Performance of base sock policies for differen Poisson disribuions when lead ime is 2. 23

24 0.1 p 0.5. Thus, E [D] = r(1 p)/p and V ar[d] = r(1 p)/p 2, leading o a variance-o-mean raio of 1/p. Table 5 compares he coss of differen policies for differen variance-o-mean raios. We observe from he able ha our base sock policies are quie robus. The relaive difference beween he opimal cos and he cos of order-up-o policies seems o be independen of he variance-o-mean raio. We observe similar resuls even for larger lead imes (no included in he paper due o space consrain), suggesing ha our policies should perform well in many pracical seings where he demand exhibis significan variance. A. Proof of Theorem 4 Proof. Since D = τ+1 =1 D, if he demand in each period has a bounded suppor, so is D. Par (a) hen follows immediaely from he definiion of m D (). To prove par (b), if he demand in each period has an IFR disribuion, i follows from Corollary 1.B.20 on page 23 in Shaked and Shanhikumar (1994) ha D also has an IFR disribuion. Then, i follows from Lemma 3 ha for any s 0, P { D 2 > s D > 2 } P { D 1 > s D > 1 }, which implies ha he mean residual life m D () is a decreasing funcion in, giving us he desired resul. To prove par (c), we can assume wihou loss of generaliy ha F (x) < 1 for all x 0. For any x, le F (x) = 1 F (x). I hen follows from he definiion of m D () ha m D () = F (u)du F. () Since E [ D 2] <, i follows ha E [D] = 0 F (u)du <. Therefore, lim F (u) = 0. Moreover, we have ha E [ D 2] = 0 2u F (u)du. Since D has a finie second momen, i follows ha lim F () = 0, implying ha boh he numeraor and he denominaor in he expression for m D ()/ converge o zero a increases o infiniy. Since D is assumed o be a coninuous random variable, we can apply 24

25 Los Negaive Binomial Variance- Opimal Bes Base Sock S B (h, b + τh) Sales Parameer o-mean Cos Level Cos % diff from Level Cos % diff from Penaly r p Raio opimal opimal % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % Table 5: Performance of base sock policies for negaive binomial disribuions wih differen variance-o-mean raio when he lead ime is 2. 25

26 L Hospial s Rule o conclude ha which is he desired resul. m D () lim = lim F () F () f() = lim 1 r() 1 = 0, B. Proof of Proposiion 7 Proof. I is easy o verify ha 1 F θ (x) = 1/(1 + x) θ. I follows ha E [D] = 0 1 F θ (x)dx = 1/(θ 1). Then, using he fac ha m D () = P {D > z} dz/p {D > }, we can also show ha m D () = (1 + )/(θ 1), which proves he firs par of Proposiion 7. To esablish he second par, noe ha by definiion S B (h, b) = F 1 θ (b/(b + h)), which implies ha S B (h, b) = ( ) b+h 1/θ h 1. Then, we have ha [ (D E S B (h, b) ) ] + = P { D > S B (h, b) } [ ] E D S B (h, b) D > S B (h, b) h = b + h m ( D S B (h, b) ) = h ( 1 + S B (h, b) ) (b + h)(θ 1), where he las equaliy follows from he formula for m D ( ). Thus, and herefore, Thus, [ (S E B (h, b) D ) ] + = E [ S B (h, b) D ] [ (D + E S B (h, b) ) ] + = S B (h, b) 1 θ 1 + h ( 1 + S B (h, b) ) (θ 1)(b + h), C B (h, b) = [ (S he B (h, b) D ) ] [ + (D + be S B (h, b) ) ] + = ( h S B (h, b) 1 ) + h ( 1 + S B (h, b) ) = hθsb (h, b). θ 1 θ 1 θ 1 which is he desired resul. C B (h, νb) lim b C B (h, b) = lim S B (h, νb) b S B (h, b) = lim b ( νb+h h ( b+h h ) 1/θ 1 ) 1/θ = ν 1/θ, 1 C. Proof of Lemma 10 Proof. Le M = sup {x : P (D x) = 0} denoe he lowes possible single period demand. Huh e al. (2006) show he convergence of he sochasic process {X L,S } for all S > M (τ + 1). This 26

Inventory Control of Perishable Items in a Two-Echelon Supply Chain

Inventory Control of Perishable Items in a Two-Echelon Supply Chain Journal of Indusrial Engineering, Universiy of ehran, Special Issue,, PP. 69-77 69 Invenory Conrol of Perishable Iems in a wo-echelon Supply Chain Fariborz Jolai *, Elmira Gheisariha and Farnaz Nojavan

More information

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB

T L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB Elecronic Companion EC.1. Proofs of Technical Lemmas and Theorems LEMMA 1. Le C(RB) be he oal cos incurred by he RB policy. Then we have, T L E[C(RB)] 3 E[Z RB ]. (EC.1) Proof of Lemma 1. Using he marginal

More information

On the Optimal Policy Structure in Serial Inventory Systems with Lost Sales

On the Optimal Policy Structure in Serial Inventory Systems with Lost Sales On he Opimal Policy Srucure in Serial Invenory Sysems wih Los Sales Woonghee Tim Huh, Columbia Universiy Ganesh Janakiraman, New York Universiy May 21, 2008 Revised: July 30, 2008; December 23, 2008 Absrac

More information

Vehicle Arrival Models : Headway

Vehicle Arrival Models : Headway Chaper 12 Vehicle Arrival Models : Headway 12.1 Inroducion Modelling arrival of vehicle a secion of road is an imporan sep in raffic flow modelling. I has imporan applicaion in raffic flow simulaion where

More information

Inventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions

Inventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions Muli-Period Sochasic Models: Opimali of (s, S) Polic for -Convex Objecive Funcions Consider a seing similar o he N-sage newsvendor problem excep ha now here is a fixed re-ordering cos (> 0) for each (re-)order.

More information

Excel-Based Solution Method For The Optimal Policy Of The Hadley And Whittin s Exact Model With Arma Demand

Excel-Based Solution Method For The Optimal Policy Of The Hadley And Whittin s Exact Model With Arma Demand Excel-Based Soluion Mehod For The Opimal Policy Of The Hadley And Whiin s Exac Model Wih Arma Demand Kal Nami School of Business and Economics Winson Salem Sae Universiy Winson Salem, NC 27110 Phone: (336)750-2338

More information

Competitive and Cooperative Inventory Policies in a Two-Stage Supply-Chain

Competitive and Cooperative Inventory Policies in a Two-Stage Supply-Chain Compeiive and Cooperaive Invenory Policies in a Two-Sage Supply-Chain (G. P. Cachon and P. H. Zipkin) Presened by Shruivandana Sharma IOE 64, Supply Chain Managemen, Winer 2009 Universiy of Michigan, Ann

More information

5. Stochastic processes (1)

5. Stochastic processes (1) Lec05.pp S-38.45 - Inroducion o Teleraffic Theory Spring 2005 Conens Basic conceps Poisson process 2 Sochasic processes () Consider some quaniy in a eleraffic (or any) sysem I ypically evolves in ime randomly

More information

Technical Report Doc ID: TR March-2013 (Last revision: 23-February-2016) On formulating quadratic functions in optimization models.

Technical Report Doc ID: TR March-2013 (Last revision: 23-February-2016) On formulating quadratic functions in optimization models. Technical Repor Doc ID: TR--203 06-March-203 (Las revision: 23-Februar-206) On formulaing quadraic funcions in opimizaion models. Auhor: Erling D. Andersen Convex quadraic consrains quie frequenl appear

More information

International Journal of Scientific & Engineering Research, Volume 4, Issue 10, October ISSN

International Journal of Scientific & Engineering Research, Volume 4, Issue 10, October ISSN Inernaional Journal of Scienific & Engineering Research, Volume 4, Issue 10, Ocober-2013 900 FUZZY MEAN RESIDUAL LIFE ORDERING OF FUZZY RANDOM VARIABLES J. EARNEST LAZARUS PIRIYAKUMAR 1, A. YAMUNA 2 1.

More information

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality

Matrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality Marix Versions of Some Refinemens of he Arihmeic-Geomeric Mean Inequaliy Bao Qi Feng and Andrew Tonge Absrac. We esablish marix versions of refinemens due o Alzer ], Carwrigh and Field 4], and Mercer 5]

More information

Robust estimation based on the first- and third-moment restrictions of the power transformation model

Robust estimation based on the first- and third-moment restrictions of the power transformation model h Inernaional Congress on Modelling and Simulaion, Adelaide, Ausralia, 6 December 3 www.mssanz.org.au/modsim3 Robus esimaion based on he firs- and hird-momen resricions of he power ransformaion Nawaa,

More information

Stochastic Perishable Inventory Systems: Dual-Balancing and Look-Ahead Approaches

Stochastic Perishable Inventory Systems: Dual-Balancing and Look-Ahead Approaches Sochasic Perishable Invenory Sysems: Dual-Balancing and Look-Ahead Approaches by Yuhe Diao A hesis presened o he Universiy Of Waerloo in fulfilmen of he hesis requiremen for he degree of Maser of Applied

More information

Final Spring 2007

Final Spring 2007 .615 Final Spring 7 Overview The purpose of he final exam is o calculae he MHD β limi in a high-bea oroidal okamak agains he dangerous n = 1 exernal ballooning-kink mode. Effecively, his corresponds o

More information

Appendix to Creating Work Breaks From Available Idleness

Appendix to Creating Work Breaks From Available Idleness Appendix o Creaing Work Breaks From Available Idleness Xu Sun and Ward Whi Deparmen of Indusrial Engineering and Operaions Research, Columbia Universiy, New York, NY, 127; {xs2235,ww24}@columbia.edu Sepember

More information

CENTRALIZED VERSUS DECENTRALIZED PRODUCTION PLANNING IN SUPPLY CHAINS

CENTRALIZED VERSUS DECENTRALIZED PRODUCTION PLANNING IN SUPPLY CHAINS CENRALIZED VERSUS DECENRALIZED PRODUCION PLANNING IN SUPPLY CHAINS Georges SAHARIDIS* a, Yves DALLERY* a, Fikri KARAESMEN* b * a Ecole Cenrale Paris Deparmen of Indusial Engineering (LGI), +3343388, saharidis,dallery@lgi.ecp.fr

More information

Lecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still.

Lecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still. Lecure - Kinemaics in One Dimension Displacemen, Velociy and Acceleraion Everyhing in he world is moving. Nohing says sill. Moion occurs a all scales of he universe, saring from he moion of elecrons in

More information

On Measuring Pro-Poor Growth. 1. On Various Ways of Measuring Pro-Poor Growth: A Short Review of the Literature

On Measuring Pro-Poor Growth. 1. On Various Ways of Measuring Pro-Poor Growth: A Short Review of the Literature On Measuring Pro-Poor Growh 1. On Various Ways of Measuring Pro-Poor Growh: A Shor eview of he Lieraure During he pas en years or so here have been various suggesions concerning he way one should check

More information

Title: Leadtime Management in a Periodic-Review Inventory System: A State-Dependent Base-Stock Policy

Title: Leadtime Management in a Periodic-Review Inventory System: A State-Dependent Base-Stock Policy Elsevier Ediorial Sysem(m) for European Journal of Operaional Research Manuscrip Draf Manuscrip Number: Tile: Leadime Managemen in a Periodic-Review Invenory Sysem: A Sae-Dependen Base-Sock Policy Aricle

More information

On Multicomponent System Reliability with Microshocks - Microdamages Type of Components Interaction

On Multicomponent System Reliability with Microshocks - Microdamages Type of Components Interaction On Mulicomponen Sysem Reliabiliy wih Microshocks - Microdamages Type of Componens Ineracion Jerzy K. Filus, and Lidia Z. Filus Absrac Consider a wo componen parallel sysem. The defined new sochasic dependences

More information

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle Chaper 2 Newonian Mechanics Single Paricle In his Chaper we will review wha Newon s laws of mechanics ell us abou he moion of a single paricle. Newon s laws are only valid in suiable reference frames,

More information

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles Diebold, Chaper 7 Francis X. Diebold, Elemens of Forecasing, 4h Ediion (Mason, Ohio: Cengage Learning, 006). Chaper 7. Characerizing Cycles Afer compleing his reading you should be able o: Define covariance

More information

Lecture 20: Riccati Equations and Least Squares Feedback Control

Lecture 20: Riccati Equations and Least Squares Feedback Control 34-5 LINEAR SYSTEMS Lecure : Riccai Equaions and Leas Squares Feedback Conrol 5.6.4 Sae Feedback via Riccai Equaions A recursive approach in generaing he marix-valued funcion W ( ) equaion for i for he

More information

An Inventory Model for Time Dependent Weibull Deterioration with Partial Backlogging

An Inventory Model for Time Dependent Weibull Deterioration with Partial Backlogging American Journal of Operaional Research 0, (): -5 OI: 0.593/j.ajor.000.0 An Invenory Model for Time ependen Weibull eerioraion wih Parial Backlogging Umakana Mishra,, Chaianya Kumar Tripahy eparmen of

More information

STATE-SPACE MODELLING. A mass balance across the tank gives:

STATE-SPACE MODELLING. A mass balance across the tank gives: B. Lennox and N.F. Thornhill, 9, Sae Space Modelling, IChemE Process Managemen and Conrol Subjec Group Newsleer STE-SPACE MODELLING Inroducion: Over he pas decade or so here has been an ever increasing

More information

Deteriorating Inventory Model with Time. Dependent Demand and Partial Backlogging

Deteriorating Inventory Model with Time. Dependent Demand and Partial Backlogging Applied Mahemaical Sciences, Vol. 4, 00, no. 7, 36-369 Deerioraing Invenory Model wih Time Dependen Demand and Parial Backlogging Vinod Kumar Mishra Deparmen of Compuer Science & Engineering Kumaon Engineering

More information

On a Discrete-In-Time Order Level Inventory Model for Items with Random Deterioration

On a Discrete-In-Time Order Level Inventory Model for Items with Random Deterioration Journal of Agriculure and Life Sciences Vol., No. ; June 4 On a Discree-In-Time Order Level Invenory Model for Iems wih Random Deerioraion Dr Biswaranjan Mandal Associae Professor of Mahemaics Acharya

More information

Stochastic models and their distributions

Stochastic models and their distributions Sochasic models and heir disribuions Couning cusomers Suppose ha n cusomers arrive a a grocery a imes, say T 1,, T n, each of which akes any real number in he inerval (, ) equally likely The values T 1,,

More information

Introduction to Probability and Statistics Slides 4 Chapter 4

Introduction to Probability and Statistics Slides 4 Chapter 4 Inroducion o Probabiliy and Saisics Slides 4 Chaper 4 Ammar M. Sarhan, asarhan@mahsa.dal.ca Deparmen of Mahemaics and Saisics, Dalhousie Universiy Fall Semeser 8 Dr. Ammar Sarhan Chaper 4 Coninuous Random

More information

EXERCISES FOR SECTION 1.5

EXERCISES FOR SECTION 1.5 1.5 Exisence and Uniqueness of Soluions 43 20. 1 v c 21. 1 v c 1 2 4 6 8 10 1 2 2 4 6 8 10 Graph of approximae soluion obained using Euler s mehod wih = 0.1. Graph of approximae soluion obained using Euler

More information

Mathematical Theory and Modeling ISSN (Paper) ISSN (Online) Vol 3, No.3, 2013

Mathematical Theory and Modeling ISSN (Paper) ISSN (Online) Vol 3, No.3, 2013 Mahemaical Theory and Modeling ISSN -580 (Paper) ISSN 5-05 (Online) Vol, No., 0 www.iise.org The ffec of Inverse Transformaion on he Uni Mean and Consan Variance Assumpions of a Muliplicaive rror Model

More information

20. Applications of the Genetic-Drift Model

20. Applications of the Genetic-Drift Model 0. Applicaions of he Geneic-Drif Model 1) Deermining he probabiliy of forming any paricular combinaion of genoypes in he nex generaion: Example: If he parenal allele frequencies are p 0 = 0.35 and q 0

More information

CHERNOFF DISTANCE AND AFFINITY FOR TRUNCATED DISTRIBUTIONS *

CHERNOFF DISTANCE AND AFFINITY FOR TRUNCATED DISTRIBUTIONS * haper 5 HERNOFF DISTANE AND AFFINITY FOR TRUNATED DISTRIBUTIONS * 5. Inroducion In he case of disribuions ha saisfy he regulariy condiions, he ramer- Rao inequaliy holds and he maximum likelihood esimaor

More information

A Shooting Method for A Node Generation Algorithm

A Shooting Method for A Node Generation Algorithm A Shooing Mehod for A Node Generaion Algorihm Hiroaki Nishikawa W.M.Keck Foundaion Laboraory for Compuaional Fluid Dynamics Deparmen of Aerospace Engineering, Universiy of Michigan, Ann Arbor, Michigan

More information

Chapter 2. First Order Scalar Equations

Chapter 2. First Order Scalar Equations Chaper. Firs Order Scalar Equaions We sar our sudy of differenial equaions in he same way he pioneers in his field did. We show paricular echniques o solve paricular ypes of firs order differenial equaions.

More information

An introduction to the theory of SDDP algorithm

An introduction to the theory of SDDP algorithm An inroducion o he heory of SDDP algorihm V. Leclère (ENPC) Augus 1, 2014 V. Leclère Inroducion o SDDP Augus 1, 2014 1 / 21 Inroducion Large scale sochasic problem are hard o solve. Two ways of aacking

More information

Explaining Total Factor Productivity. Ulrich Kohli University of Geneva December 2015

Explaining Total Factor Productivity. Ulrich Kohli University of Geneva December 2015 Explaining Toal Facor Produciviy Ulrich Kohli Universiy of Geneva December 2015 Needed: A Theory of Toal Facor Produciviy Edward C. Presco (1998) 2 1. Inroducion Toal Facor Produciviy (TFP) has become

More information

Let us start with a two dimensional case. We consider a vector ( x,

Let us start with a two dimensional case. We consider a vector ( x, Roaion marices We consider now roaion marices in wo and hree dimensions. We sar wih wo dimensions since wo dimensions are easier han hree o undersand, and one dimension is a lile oo simple. However, our

More information

Application of a Stochastic-Fuzzy Approach to Modeling Optimal Discrete Time Dynamical Systems by Using Large Scale Data Processing

Application of a Stochastic-Fuzzy Approach to Modeling Optimal Discrete Time Dynamical Systems by Using Large Scale Data Processing Applicaion of a Sochasic-Fuzzy Approach o Modeling Opimal Discree Time Dynamical Sysems by Using Large Scale Daa Processing AA WALASZE-BABISZEWSA Deparmen of Compuer Engineering Opole Universiy of Technology

More information

The Strong Law of Large Numbers

The Strong Law of Large Numbers Lecure 9 The Srong Law of Large Numbers Reading: Grimme-Sirzaker 7.2; David Williams Probabiliy wih Maringales 7.2 Furher reading: Grimme-Sirzaker 7.1, 7.3-7.5 Wih he Convergence Theorem (Theorem 54) and

More information

MATH 4330/5330, Fourier Analysis Section 6, Proof of Fourier s Theorem for Pointwise Convergence

MATH 4330/5330, Fourier Analysis Section 6, Proof of Fourier s Theorem for Pointwise Convergence MATH 433/533, Fourier Analysis Secion 6, Proof of Fourier s Theorem for Poinwise Convergence Firs, some commens abou inegraing periodic funcions. If g is a periodic funcion, g(x + ) g(x) for all real x,

More information

Lecture 4 Notes (Little s Theorem)

Lecture 4 Notes (Little s Theorem) Lecure 4 Noes (Lile s Theorem) This lecure concerns one of he mos imporan (and simples) heorems in Queuing Theory, Lile s Theorem. More informaion can be found in he course book, Bersekas & Gallagher,

More information

5.1 - Logarithms and Their Properties

5.1 - Logarithms and Their Properties Chaper 5 Logarihmic Funcions 5.1 - Logarihms and Their Properies Suppose ha a populaion grows according o he formula P 10, where P is he colony size a ime, in hours. When will he populaion be 2500? We

More information

3.1.3 INTRODUCTION TO DYNAMIC OPTIMIZATION: DISCRETE TIME PROBLEMS. A. The Hamiltonian and First-Order Conditions in a Finite Time Horizon

3.1.3 INTRODUCTION TO DYNAMIC OPTIMIZATION: DISCRETE TIME PROBLEMS. A. The Hamiltonian and First-Order Conditions in a Finite Time Horizon 3..3 INRODUCION O DYNAMIC OPIMIZAION: DISCREE IME PROBLEMS A. he Hamilonian and Firs-Order Condiions in a Finie ime Horizon Define a new funcion, he Hamilonian funcion, H. H he change in he oal value of

More information

Seminar 4: Hotelling 2

Seminar 4: Hotelling 2 Seminar 4: Hoelling 2 November 3, 211 1 Exercise Par 1 Iso-elasic demand A non renewable resource of a known sock S can be exraced a zero cos. Demand for he resource is of he form: D(p ) = p ε ε > A a

More information

Cash Flow Valuation Mode Lin Discrete Time

Cash Flow Valuation Mode Lin Discrete Time IOSR Journal of Mahemaics (IOSR-JM) e-issn: 2278-5728,p-ISSN: 2319-765X, 6, Issue 6 (May. - Jun. 2013), PP 35-41 Cash Flow Valuaion Mode Lin Discree Time Olayiwola. M. A. and Oni, N. O. Deparmen of Mahemaics

More information

Maintenance Models. Prof. Robert C. Leachman IEOR 130, Methods of Manufacturing Improvement Spring, 2011

Maintenance Models. Prof. Robert C. Leachman IEOR 130, Methods of Manufacturing Improvement Spring, 2011 Mainenance Models Prof Rober C Leachman IEOR 3, Mehods of Manufacuring Improvemen Spring, Inroducion The mainenance of complex equipmen ofen accouns for a large porion of he coss associaed wih ha equipmen

More information

A Study of Inventory System with Ramp Type Demand Rate and Shortage in The Light Of Inflation I

A Study of Inventory System with Ramp Type Demand Rate and Shortage in The Light Of Inflation I Inernaional Journal of Mahemaics rends and echnology Volume 7 Number Jan 5 A Sudy of Invenory Sysem wih Ramp ype emand Rae and Shorage in he Ligh Of Inflaion I Sangeea Gupa, R.K. Srivasava, A.K. Singh

More information

Stability and Bifurcation in a Neural Network Model with Two Delays

Stability and Bifurcation in a Neural Network Model with Two Delays Inernaional Mahemaical Forum, Vol. 6, 11, no. 35, 175-1731 Sabiliy and Bifurcaion in a Neural Nework Model wih Two Delays GuangPing Hu and XiaoLing Li School of Mahemaics and Physics, Nanjing Universiy

More information

The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales

The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales Advances in Dynamical Sysems and Applicaions. ISSN 0973-5321 Volume 1 Number 1 (2006, pp. 103 112 c Research India Publicaions hp://www.ripublicaion.com/adsa.hm The Asympoic Behavior of Nonoscillaory Soluions

More information

Optimality Conditions for Unconstrained Problems

Optimality Conditions for Unconstrained Problems 62 CHAPTER 6 Opimaliy Condiions for Unconsrained Problems 1 Unconsrained Opimizaion 11 Exisence Consider he problem of minimizing he funcion f : R n R where f is coninuous on all of R n : P min f(x) x

More information

Applying Genetic Algorithms for Inventory Lot-Sizing Problem with Supplier Selection under Storage Capacity Constraints

Applying Genetic Algorithms for Inventory Lot-Sizing Problem with Supplier Selection under Storage Capacity Constraints IJCSI Inernaional Journal of Compuer Science Issues, Vol 9, Issue 1, No 1, January 2012 wwwijcsiorg 18 Applying Geneic Algorihms for Invenory Lo-Sizing Problem wih Supplier Selecion under Sorage Capaciy

More information

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H.

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H. ACE 56 Fall 005 Lecure 5: he Simple Linear Regression Model: Sampling Properies of he Leas Squares Esimaors by Professor Sco H. Irwin Required Reading: Griffihs, Hill and Judge. "Inference in he Simple

More information

Approximation algorithms for capacitated stochastic inventory systems with setup costs

Approximation algorithms for capacitated stochastic inventory systems with setup costs Approximaion algorihms for capaciaed sochasic invenory sysems wih seup coss The MIT Faculy has made his aricle openly available. Please share how his access benefis you. Your sory maers. Ciaion As Published

More information

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems

On Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems MATHEMATICS OF OPERATIONS RESEARCH Vol. 38, No. 2, May 2013, pp. 209 227 ISSN 0364-765X (prin) ISSN 1526-5471 (online) hp://dx.doi.org/10.1287/moor.1120.0562 2013 INFORMS On Boundedness of Q-Learning Ieraes

More information

Notes for Lecture 17-18

Notes for Lecture 17-18 U.C. Berkeley CS278: Compuaional Complexiy Handou N7-8 Professor Luca Trevisan April 3-8, 2008 Noes for Lecure 7-8 In hese wo lecures we prove he firs half of he PCP Theorem, he Amplificaion Lemma, up

More information

Simulation-Solving Dynamic Models ABE 5646 Week 2, Spring 2010

Simulation-Solving Dynamic Models ABE 5646 Week 2, Spring 2010 Simulaion-Solving Dynamic Models ABE 5646 Week 2, Spring 2010 Week Descripion Reading Maerial 2 Compuer Simulaion of Dynamic Models Finie Difference, coninuous saes, discree ime Simple Mehods Euler Trapezoid

More information

Solutions Problem Set 3 Macro II (14.452)

Solutions Problem Set 3 Macro II (14.452) Soluions Problem Se 3 Macro II (14.452) Francisco A. Gallego 04/27/2005 1 Q heory of invesmen in coninuous ime and no uncerainy Consider he in nie horizon model of a rm facing adjusmen coss o invesmen.

More information

1 Review of Zero-Sum Games

1 Review of Zero-Sum Games COS 5: heoreical Machine Learning Lecurer: Rob Schapire Lecure #23 Scribe: Eugene Brevdo April 30, 2008 Review of Zero-Sum Games Las ime we inroduced a mahemaical model for wo player zero-sum games. Any

More information

18 Biological models with discrete time

18 Biological models with discrete time 8 Biological models wih discree ime The mos imporan applicaions, however, may be pedagogical. The elegan body of mahemaical heory peraining o linear sysems (Fourier analysis, orhogonal funcions, and so

More information

On a Fractional Stochastic Landau-Ginzburg Equation

On a Fractional Stochastic Landau-Ginzburg Equation Applied Mahemaical Sciences, Vol. 4, 1, no. 7, 317-35 On a Fracional Sochasic Landau-Ginzburg Equaion Nguyen Tien Dung Deparmen of Mahemaics, FPT Universiy 15B Pham Hung Sree, Hanoi, Vienam dungn@fp.edu.vn

More information

Chapter 2. Models, Censoring, and Likelihood for Failure-Time Data

Chapter 2. Models, Censoring, and Likelihood for Failure-Time Data Chaper 2 Models, Censoring, and Likelihood for Failure-Time Daa William Q. Meeker and Luis A. Escobar Iowa Sae Universiy and Louisiana Sae Universiy Copyrigh 1998-2008 W. Q. Meeker and L. A. Escobar. Based

More information

Incorporating Delay Mechanism in Ordering Policies in Multi- Echelon Distribution Systems

Incorporating Delay Mechanism in Ordering Policies in Multi- Echelon Distribution Systems Incorporaing Delay Mechanism in Ordering Policies in Muli- Echelon Disribuion Sysems Kamran Moinzadeh and Yong-Pin Zhou Universiy of Washingon Business School {amran, yongpin}@u.washingon.edu June, 2005;

More information

Online Appendix to Solution Methods for Models with Rare Disasters

Online Appendix to Solution Methods for Models with Rare Disasters Online Appendix o Soluion Mehods for Models wih Rare Disasers Jesús Fernández-Villaverde and Oren Levinal In his Online Appendix, we presen he Euler condiions of he model, we develop he pricing Calvo block,

More information

Errata (1 st Edition)

Errata (1 st Edition) P Sandborn, os Analysis of Elecronic Sysems, s Ediion, orld Scienific, Singapore, 03 Erraa ( s Ediion) S K 05D Page 8 Equaion (7) should be, E 05D E Nu e S K he L appearing in he equaion in he book does

More information

not to be republished NCERT MATHEMATICAL MODELLING Appendix 2 A.2.1 Introduction A.2.2 Why Mathematical Modelling?

not to be republished NCERT MATHEMATICAL MODELLING Appendix 2 A.2.1 Introduction A.2.2 Why Mathematical Modelling? 256 MATHEMATICS A.2.1 Inroducion In class XI, we have learn abou mahemaical modelling as an aemp o sudy some par (or form) of some real-life problems in mahemaical erms, i.e., he conversion of a physical

More information

ACE 564 Spring Lecture 7. Extensions of The Multiple Regression Model: Dummy Independent Variables. by Professor Scott H.

ACE 564 Spring Lecture 7. Extensions of The Multiple Regression Model: Dummy Independent Variables. by Professor Scott H. ACE 564 Spring 2006 Lecure 7 Exensions of The Muliple Regression Model: Dumm Independen Variables b Professor Sco H. Irwin Readings: Griffihs, Hill and Judge. "Dumm Variables and Varing Coefficien Models

More information

Approximation Algorithms for Unique Games via Orthogonal Separators

Approximation Algorithms for Unique Games via Orthogonal Separators Approximaion Algorihms for Unique Games via Orhogonal Separaors Lecure noes by Konsanin Makarychev. Lecure noes are based on he papers [CMM06a, CMM06b, LM4]. Unique Games In hese lecure noes, we define

More information

IMPLICIT AND INVERSE FUNCTION THEOREMS PAUL SCHRIMPF 1 OCTOBER 25, 2013

IMPLICIT AND INVERSE FUNCTION THEOREMS PAUL SCHRIMPF 1 OCTOBER 25, 2013 IMPLICI AND INVERSE FUNCION HEOREMS PAUL SCHRIMPF 1 OCOBER 25, 213 UNIVERSIY OF BRIISH COLUMBIA ECONOMICS 526 We have exensively sudied how o solve sysems of linear equaions. We know how o check wheher

More information

A Hop Constrained Min-Sum Arborescence with Outage Costs

A Hop Constrained Min-Sum Arborescence with Outage Costs A Hop Consrained Min-Sum Arborescence wih Ouage Coss Rakesh Kawara Minnesoa Sae Universiy, Mankao, MN 56001 Email: Kawara@mnsu.edu Absrac The hop consrained min-sum arborescence wih ouage coss problem

More information

10. State Space Methods

10. State Space Methods . Sae Space Mehods. Inroducion Sae space modelling was briefly inroduced in chaper. Here more coverage is provided of sae space mehods before some of heir uses in conrol sysem design are covered in he

More information

15. Vector Valued Functions

15. Vector Valued Functions 1. Vecor Valued Funcions Up o his poin, we have presened vecors wih consan componens, for example, 1, and,,4. However, we can allow he componens of a vecor o be funcions of a common variable. For example,

More information

Probabilistic Models for Reliability Analysis of a System with Three Consecutive Stages of Deterioration

Probabilistic Models for Reliability Analysis of a System with Three Consecutive Stages of Deterioration Yusuf I., Gaawa R.I. Volume, December 206 Probabilisic Models for Reliabiliy Analysis of a Sysem wih Three Consecuive Sages of Deerioraion Ibrahim Yusuf Deparmen of Mahemaical Sciences, Bayero Universiy,

More information

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes Represening Periodic Funcions by Fourier Series 3. Inroducion In his Secion we show how a periodic funcion can be expressed as a series of sines and cosines. We begin by obaining some sandard inegrals

More information

Deteriorating Inventory Model When Demand Depends on Advertisement and Stock Display

Deteriorating Inventory Model When Demand Depends on Advertisement and Stock Display Inernaional Journal of Operaions Research Inernaional Journal of Operaions Research Vol. 6, No. 2, 33 44 (29) Deerioraing Invenory Model When Demand Depends on Adverisemen and Sock Display Nia H. Shah,

More information

arxiv: v1 [math.pr] 19 Feb 2011

arxiv: v1 [math.pr] 19 Feb 2011 A NOTE ON FELLER SEMIGROUPS AND RESOLVENTS VADIM KOSTRYKIN, JÜRGEN POTTHOFF, AND ROBERT SCHRADER ABSTRACT. Various equivalen condiions for a semigroup or a resolven generaed by a Markov process o be of

More information

CHAPTER 2 Signals And Spectra

CHAPTER 2 Signals And Spectra CHAPER Signals And Specra Properies of Signals and Noise In communicaion sysems he received waveform is usually caegorized ino he desired par conaining he informaion, and he undesired par. he desired par

More information

Macroeconomic Theory Ph.D. Qualifying Examination Fall 2005 ANSWER EACH PART IN A SEPARATE BLUE BOOK. PART ONE: ANSWER IN BOOK 1 WEIGHT 1/3

Macroeconomic Theory Ph.D. Qualifying Examination Fall 2005 ANSWER EACH PART IN A SEPARATE BLUE BOOK. PART ONE: ANSWER IN BOOK 1 WEIGHT 1/3 Macroeconomic Theory Ph.D. Qualifying Examinaion Fall 2005 Comprehensive Examinaion UCLA Dep. of Economics You have 4 hours o complee he exam. There are hree pars o he exam. Answer all pars. Each par has

More information

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits DOI: 0.545/mjis.07.5009 Exponenial Weighed Moving Average (EWMA) Char Under The Assumpion of Moderaeness And Is 3 Conrol Limis KALPESH S TAILOR Assisan Professor, Deparmen of Saisics, M. K. Bhavnagar Universiy,

More information

Supplement for Stochastic Convex Optimization: Faster Local Growth Implies Faster Global Convergence

Supplement for Stochastic Convex Optimization: Faster Local Growth Implies Faster Global Convergence Supplemen for Sochasic Convex Opimizaion: Faser Local Growh Implies Faser Global Convergence Yi Xu Qihang Lin ianbao Yang Proof of heorem heorem Suppose Assumpion holds and F (w) obeys he LGC (6) Given

More information

6.2 Transforms of Derivatives and Integrals.

6.2 Transforms of Derivatives and Integrals. SEC. 6.2 Transforms of Derivaives and Inegrals. ODEs 2 3 33 39 23. Change of scale. If l( f ()) F(s) and c is any 33 45 APPLICATION OF s-shifting posiive consan, show ha l( f (c)) F(s>c)>c (Hin: In Probs.

More information

Christos Papadimitriou & Luca Trevisan November 22, 2016

Christos Papadimitriou & Luca Trevisan November 22, 2016 U.C. Bereley CS170: Algorihms Handou LN-11-22 Chrisos Papadimiriou & Luca Trevisan November 22, 2016 Sreaming algorihms In his lecure and he nex one we sudy memory-efficien algorihms ha process a sream

More information

Expert Advice for Amateurs

Expert Advice for Amateurs Exper Advice for Amaeurs Ernes K. Lai Online Appendix - Exisence of Equilibria The analysis in his secion is performed under more general payoff funcions. Wihou aking an explici form, he payoffs of he

More information

Homework 4 (Stats 620, Winter 2017) Due Tuesday Feb 14, in class Questions are derived from problems in Stochastic Processes by S. Ross.

Homework 4 (Stats 620, Winter 2017) Due Tuesday Feb 14, in class Questions are derived from problems in Stochastic Processes by S. Ross. Homework 4 (Sas 62, Winer 217) Due Tuesday Feb 14, in class Quesions are derived from problems in Sochasic Processes by S. Ross. 1. Le A() and Y () denoe respecively he age and excess a. Find: (a) P{Y

More information

Module 2 F c i k c s la l w a s o s f dif di fusi s o i n

Module 2 F c i k c s la l w a s o s f dif di fusi s o i n Module Fick s laws of diffusion Fick s laws of diffusion and hin film soluion Adolf Fick (1855) proposed: d J α d d d J (mole/m s) flu (m /s) diffusion coefficien and (mole/m 3 ) concenraion of ions, aoms

More information

Article from. Predictive Analytics and Futurism. July 2016 Issue 13

Article from. Predictive Analytics and Futurism. July 2016 Issue 13 Aricle from Predicive Analyics and Fuurism July 6 Issue An Inroducion o Incremenal Learning By Qiang Wu and Dave Snell Machine learning provides useful ools for predicive analyics The ypical machine learning

More information

Ann. Funct. Anal. 2 (2011), no. 2, A nnals of F unctional A nalysis ISSN: (electronic) URL:

Ann. Funct. Anal. 2 (2011), no. 2, A nnals of F unctional A nalysis ISSN: (electronic) URL: Ann. Func. Anal. 2 2011, no. 2, 34 41 A nnals of F uncional A nalysis ISSN: 2008-8752 elecronic URL: www.emis.de/journals/afa/ CLASSIFICAION OF POSIIVE SOLUIONS OF NONLINEAR SYSEMS OF VOLERRA INEGRAL EQUAIONS

More information

A Primal-Dual Type Algorithm with the O(1/t) Convergence Rate for Large Scale Constrained Convex Programs

A Primal-Dual Type Algorithm with the O(1/t) Convergence Rate for Large Scale Constrained Convex Programs PROC. IEEE CONFERENCE ON DECISION AND CONTROL, 06 A Primal-Dual Type Algorihm wih he O(/) Convergence Rae for Large Scale Consrained Convex Programs Hao Yu and Michael J. Neely Absrac This paper considers

More information

Bernoulli numbers. Francesco Chiatti, Matteo Pintonello. December 5, 2016

Bernoulli numbers. Francesco Chiatti, Matteo Pintonello. December 5, 2016 UNIVERSITÁ DEGLI STUDI DI PADOVA, DIPARTIMENTO DI MATEMATICA TULLIO LEVI-CIVITA Bernoulli numbers Francesco Chiai, Maeo Pinonello December 5, 206 During las lessons we have proved he Las Ferma Theorem

More information

14 Autoregressive Moving Average Models

14 Autoregressive Moving Average Models 14 Auoregressive Moving Average Models In his chaper an imporan parameric family of saionary ime series is inroduced, he family of he auoregressive moving average, or ARMA, processes. For a large class

More information

Sensors, Signals and Noise

Sensors, Signals and Noise Sensors, Signals and Noise COURSE OUTLINE Inroducion Signals and Noise: 1) Descripion Filering Sensors and associaed elecronics rv 2017/02/08 1 Noise Descripion Noise Waveforms and Samples Saisics of Noise

More information

Two Coupled Oscillators / Normal Modes

Two Coupled Oscillators / Normal Modes Lecure 3 Phys 3750 Two Coupled Oscillaors / Normal Modes Overview and Moivaion: Today we ake a small, bu significan, sep owards wave moion. We will no ye observe waves, bu his sep is imporan in is own

More information

Energy Storage Benchmark Problems

Energy Storage Benchmark Problems Energy Sorage Benchmark Problems Daniel F. Salas 1,3, Warren B. Powell 2,3 1 Deparmen of Chemical & Biological Engineering 2 Deparmen of Operaions Research & Financial Engineering 3 Princeon Laboraory

More information

Hamilton- J acobi Equation: Explicit Formulas In this lecture we try to apply the method of characteristics to the Hamilton-Jacobi equation: u t

Hamilton- J acobi Equation: Explicit Formulas In this lecture we try to apply the method of characteristics to the Hamilton-Jacobi equation: u t M ah 5 2 7 Fall 2 0 0 9 L ecure 1 0 O c. 7, 2 0 0 9 Hamilon- J acobi Equaion: Explici Formulas In his lecure we ry o apply he mehod of characerisics o he Hamilon-Jacobi equaion: u + H D u, x = 0 in R n

More information

Comparing Means: t-tests for One Sample & Two Related Samples

Comparing Means: t-tests for One Sample & Two Related Samples Comparing Means: -Tess for One Sample & Two Relaed Samples Using he z-tes: Assumpions -Tess for One Sample & Two Relaed Samples The z-es (of a sample mean agains a populaion mean) is based on he assumpion

More information

Reliability of Technical Systems

Reliability of Technical Systems eliabiliy of Technical Sysems Main Topics Inroducion, Key erms, framing he problem eliabiliy parameers: Failure ae, Failure Probabiliy, Availabiliy, ec. Some imporan reliabiliy disribuions Componen reliabiliy

More information

Lecture Notes 2. The Hilbert Space Approach to Time Series

Lecture Notes 2. The Hilbert Space Approach to Time Series Time Series Seven N. Durlauf Universiy of Wisconsin. Basic ideas Lecure Noes. The Hilber Space Approach o Time Series The Hilber space framework provides a very powerful language for discussing he relaionship

More information

Class Meeting # 10: Introduction to the Wave Equation

Class Meeting # 10: Introduction to the Wave Equation MATH 8.5 COURSE NOTES - CLASS MEETING # 0 8.5 Inroducion o PDEs, Fall 0 Professor: Jared Speck Class Meeing # 0: Inroducion o he Wave Equaion. Wha is he wave equaion? The sandard wave equaion for a funcion

More information

Two Popular Bayesian Estimators: Particle and Kalman Filters. McGill COMP 765 Sept 14 th, 2017

Two Popular Bayesian Estimators: Particle and Kalman Filters. McGill COMP 765 Sept 14 th, 2017 Two Popular Bayesian Esimaors: Paricle and Kalman Filers McGill COMP 765 Sep 14 h, 2017 1 1 1, dx x Bel x u x P x z P Recall: Bayes Filers,,,,,,, 1 1 1 1 u z u x P u z u x z P Bayes z = observaion u =

More information

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation

Course Notes for EE227C (Spring 2018): Convex Optimization and Approximation Course Noes for EE7C Spring 018: Convex Opimizaion and Approximaion Insrucor: Moriz Hard Email: hard+ee7c@berkeley.edu Graduae Insrucor: Max Simchowiz Email: msimchow+ee7c@berkeley.edu Ocober 15, 018 3

More information