Backorder minimization in multiproduct assemble-to-order systems

Size: px
Start display at page:

Download "Backorder minimization in multiproduct assemble-to-order systems"

Transcription

1 IIE Tranacton (2005) 37, Copyrght C IIE ISSN: X prnt / onlne DOI: / Backorder mnmzaton n multproduct aemble-to-order ytem YINGDONG LU, 1 JING-SHENG SONG 2 and DAVID D. YAO 3 1 IBM T.J. Waton Reearch Center, Yorktown Heght, NY 10598, USA E-mal: yngdong@u.bm.com 2 The Fuqua School of Bune, Duke Unverty, Durham, NC 27708, USA E-mal: jong@duke.edu 3 Department of Indutral Engneerng and Operaton Reearch, Columba Unverty, New York, NY 10027, USA E-mal: yao@columba.edu Receved November 2002 and accepted December 2004 We conder a multproduct aemble-to-order ytem. Component are bult to tock wth nventory controlled by bae-tock rule, but the fnal product are aembled to order. Cutomer order of each product follow a batch Poon proce. The leadtme for replenhng component nventory are tochatc. We tudy the optmal allocaton of a gven budget among component nventore o a to mnmze a weghted average of backorder over product type. We derve eay-to-compute bound and approxmaton for the expected number of backorder and ue them to formulate urrogate optmzaton problem. Effcent algorthm are developed to olve thee problem, and numercal example llutrate the effectvene of the bound and approxmaton. 1. Introducton Realzng the compettve advantage of beng flexble, reponve a well a effcent, and enabled by advanced technologe, more and more manufacturng compane are reengneerng ther product degn and delvery procee to move toward ma cutomzaton (Pne, 1993). Th requre modularzng the producton proce o that the product can be quckly aembled from tandardzed component and module n dfferent confguraton, baed on what cutomer ndvdually requet. A a reult, a new knd of producton-nventory ytem ha emerged and becomng ncreangly popular. Th the Aemble-To-Order (ATO) ytem: nventore are kept only at the component level, and the fnal product are aembled only after cutomer order are receved. Th new ytem, n turn, preent challengng operatonal and ytem degn ue to manager. A each cutomer order typcally nvolve everal component n dfferent amount, the tockout of any component wll caue a delay n fulfllng the order. So, the optmal tock level of one component hould be determned n conjuncton wth thoe of other component to enure ther multaneou avalablty. Standard ngle-tem nventory plannng tool, whle utable for the ma-producton make-to-tock envronment, are no longer applcable. New plannng tool are needed to trke the optmal nventory-ervce tradeoff n ATO ytem. The current paper preent an effort toward th goal. More pecfcally, we conder an ATO ytem upportng multple type of demand, whch arrve at the ytem followng compound Poon procee. The component nventore are reuppled from outde uppler after random replenhment leadtme. For a gven component, the leadtme are ndependent, dentcally dtrbuted (..d.) random varable. The leadtme for dfferent component are alo ndependent but may have dfferent dtrbuton. Snce the form of the optmal nventory-control polcy for th ytem unknown, bae-tock polce are wdely adopted n practce. For th reaon, we aume the nventory of each component controlled by a bae-tock polcy. The ytem performance meaure we focu on the expected backorder for each product. Our objectve to mnmze a weghted average of backorder over all product type, ubject to a budget contrant on the component nventory. Through Lttle law (Wolff, 1989), th objectve relate drectly to the repone tme performance n fulfllng cutomer order. The optmzaton problem under tudy qute complex. Frt, t objectve functon noneparable and nondfferentable. Second, t evaluaton nvolve jont probablte, whch can be computatonally challengng. To deal wth the econd dffculty, we develop upper and lower bound on the backorder that nvolve margnal dtrbuton only and ue thee bound and approxmaton X C 2005 IIE

2 764 Lu et al. derved from them a urrogate objectve functon n the optmzaton problem. Th approach mlar n prt to that n Song and Yao (2002) for the ngle-product ATO ytem. However, even wth the mpler urrogate objectve functon the frt dffculty reman. Moreover, the bound for the multproduct model here are more nvolved than n the ngle-product cae and exhbt dfferent tructure, o the dea ued to develop the algorthm n Song and Yao do not apply. In partcular, the lower bound n the multproduct cae lead to an optmzaton problem known a the mn-maxum reource allocaton problem (ouvel et al., 2001), for whch known algorthm are lmted to the branch-andbound type. One man contrbuton of th paper the dentfcaton of a pecal property n our model, whch we call a tack tructure. Ung th property, we are able to develop a much more effcent greedy algorthm to olve the problem. Another nteretng fndng that a modfcaton of the upper bound approach n Song and Yao (2002) produce not only a good approxmaton of the average backorder but alo a urrogate optmzaton problem that bear the ame tack tructure a the one n the lowerbound approach. Several other author have worked on optmzaton model for ATO ytem. Thee tude are qute dfferent from our n the detaled modelng aumpton and oluton approache. For example, ntead of the contnuou-tme cheme condered here, one tream of reearch employ a dcrete-tme (perodc-revew) model. Wthn th framework, Hauman et al. (1998) aume that demand n dfferent tme perod are ndependent, and demand n each perod follow a multvarate normal dtrbuton. Component replenhment leadtme are contant. The paper develop heurtc method to olve the problem of maxmzng a lower bound on the fll rate, whch a multvarate normal probablty, ubject to a lnear budget contrant. Ung the ame framework, Agrawal and Cohen (2001) tudy the problem of mnmzng the total expected component nventory cot ubject to contrant on the order fll rate. Intead of a frt-come frt-erved (FCFS) allocaton rule, they aume a combnaton of FCFS and a far-hare allocaton polcy. Cheng et al. (2002) alo ue the dcrete-tme formulaton and tudy the problem of mnmzng average component nventory holdng cot ubject to productfamly-dependent fll rate contrant. They aume..d. replenhment leadtme and a FCFS allocaton rule. An exact algorthm and a greedy heurtc are developed. Among artcle ung a contnuou-tme framework, Gallen and Wen (2001) aume..d. component leadtme n a ngle-product aembly ytem. Demand Poon wth rate λ. To keep the analy tractable, the paper mpoe a ynchronzaton aumpton, namely, that component are aembled n the ame equence n whch they are ordered. (Order ynchronzaton not aumed n our model.) In the context of a repar hop, Cheung and Hauman (1995) aume..d. leadtme and a multvarate Poon demand model. Th a pecal cae of our (the unt demand cae). Unlke n our paper, they aume complete component cannbalzaton, whch make ene n the repar hop but not n ATO ytem. Under th aumpton, the product backorder are the maxmum of the component backorder, o ther backorder functon much mpler than and completely dfferent from our. Lu and Song (2002) tudy the multproduct unt demand ytem. Aumng there a backorder cot rate aocated wth each backlogged cutomer order, they formulate an uncontraned optmzaton model to mnmze the expected average holdng and backorder cot. Wang (1999) conder multvarate compound Poon demand but the upply proce for each component capactated and modeled a a ngle-erver queue. Applyng the aymptotc reult developed n Glaerman and Wang (1998), he examne the problem of mnmzng average nventory cot ubject to a fll-rate contrant. Our tudy appear to be the frt attempt to tackle optmzaton of multproduct backorder wth batch demand. We refer the reader to Song and Zpkn (2003) for a more detaled urvey of the tate-of-the-art reearch on ATO ytem. The ret of the paper organzed a follow. We tart wth model decrpton and ome prelmnare n Secton 2. In Secton 3, we develop the lower bound for backorder, a well a oluton approache to the lower-bound problem. We then turn to conderng upper bound n Secton 4. In Secton 5 we preent numercal tude to ae and compare the effectvene of the oluton approache developed n the prevou ecton. 2. Model and prelmnare Let I ={1, 2,...,m} denote the et of all component ndce. Cutomer order arrve at the ytem followng a tatonary Poon proce, denoted {A(t), t 0}, wth rate λ. Each order may requre everal component n dfferent amount multaneouly. For any ubet of component I,weay an order of type f t cont of potve unt of each component n and zero unt n I\. We aume that there a fxed probablty q that an order of type, q = 1. Thu, the type- order tream form acompound Poon proce wth rate λ = q λ. A type- order requet Q j unt of component j, and Q = (Q j, j ) follow a known dcrete probablty dtrbuton. We aume that each order type ndependent of the other order type and of all other event. We denote to be the et of all demand type, that, ={ I : q > 0}. Note that not necearly the et of all poble ubet of I. Th demand model qute general and cover a number of mportant pecal cae. For example, f Q j are determntc, then Q repreent the Bll-Of-Materal (BOM) of the type- product. If Q j = Qξj where ξj are contant

3 Backorder mnmzaton n multproduct aemble-to-order ytem 765 and Q a random varable, then (ξj )repreent the BOM of the type- product wherea Q the random quantty of th product that a cutomer demand. For each component, let denote the famly of ubet of that contan. Itclear that the demand proce for component form acompound Poon proce wth rate: λ = λ, and batch ze Q, the mxture of Q for all. (Throughout the paper we ue ubcrpt to ndcate component type and upercrpt for demand type.) The overall demand rate for component thu: µ = λ E ( Q ). Note, however, that thee compound Poon procee are no longer ndependent. Demand are flled on a FCFS ba. Upon the arrval of any demand, f there enough on-hand nventory for all the component requred by the demand, then the demand flled mmedately. In other word, the tme to aemble the component nto the end-product aumed to be neglgble. We alo aume complete backloggng for demand that cannot be flled mmedately. When a demand arrve and ome of t requred component are n tock but other are not, we can ether hp the n-tock component or put them ade a commtted nventory. However, a demand condered to be backlogged unle t can be atfed completely. When there are backorder, they are alo flled on a FCFS ba. We remark that FCFS a uboptmal allocaton rule. Among other thng, an optmal allocaton polcy need to take nto account the multaneou avalablty of component tock. However, due to the complexty of the ytem, the tructure of optmal or near-optmal allocaton rule not yet known. Becaue of t mplcty and eae of mplementaton, FCFS wdely ued acro ndutre. FCFS alo make the etmaton of product backorder more tractable. Alternatve approache have been dcued n the lterature, motly n a dcrete-tme formulaton, o that ome knd of myopc allocaton ued wthn each perod. When dealng wth order n dfferent perod, however, mot of thee approache ue FCFS. In our contnuou-tme ettng, there no order batchng, o myopc allocaton not applcable. See Song and Zpkn (2003) for detal. The nventory of each component controlled by an ndependent bae-tock polcy, wth: := the bae-tock level for component. That, upon each demand arrval, f the nventory poton (.e., the on-hand nventory plu on-order poton mnu backorder) of component le than, then order up to ; otherwe, do not order. Th type of polcy known to be optmal when there are no econome of cale n replenhment and when the component demand are ndependent. When component demand are correlated a n our cae, the optmal polcy hould be a coordnated polcy. Unfortunately, the prece form of the optmal polcy tll unknown. Hence, the ndependent bae-tock polce are wdely ued n practce becaue of ther mplcty. Snce we follow a bae-tock polcy and the demand arrve n batche, each component replenhment order compre everal unt. We aume that, for each unt of component, the replenhment leadtme are..d. random varable wth a common cumulatve dtrbuton G. Let L denote the generc random varable wth dtrbuton G and mean E [L = l. Denote G c = 1 G. Aume the leadtme are ndependent among the component; that, L ndependent of L j for any j. Forany tme t, let I (t)bethe net nventory of component at t, and X (t) bethe number of outtandng order of component at t. Then, by the nature of the bae-tock control, we have: I (t) = X (t), = 1,...,m. (1) It clear that X (t)equal to the number of job n ervce n an M/G / queue, wth arrval rate λ,for = 1,...,m. However, nce the arrval of any demand of type- generate multaneou arrval at all the component queue,, thee m queue are not ndependent. Lu et al. (2003) how that the jont dtrbuton of (X 1 (t),...,x m (t)) ha a teady-tate lmt (X 1,...,X m ) wth probablty generatng functon: ψ(τ 1,...,τ m ) [ m := E j=1 τ X j j [ = exp λ ( ( ψ Q G1 (u) + τ 1 G c 1 (u),...,g m(u) + τ m G c m (u)) 1 ) du, (2) where ψ Q (x 1,...,x m ) denote the generatng functon of Q = (Q ). We can ue ψ to derve the moment and covarance of the X. In partcular: θ := E [X = l λ E ( Q ) = µ l, (3) σ 2 := Var[X = θ + λ [ E (( Q ) 2 ) E ( Q ) [ G c j (u) 2 du. (4) 0 And for j, σ j := Cov[X, X j = λ E ( Q Q ) j G c (u)gc j (u)du. :,j 0 0 (5)

4 766 Lu et al. It can be verfed that σ 2 θ for all. Alo, σ j 0 and t zero f and only f there no demand type that requre both and j. For unt demand model,.e., Q j = 1, for all j, ψ(τ 1,...,τ m )the probablty generatng functon of a multvarate Poon dtrbuton. That, the margnal dtrbuton of X Poon wth mean λ l, = 1,...,m. In th paper, the performance meaure of prmary concern are: B = teady-tate number of backordered type- demand, for all. We now how how thee quantte depend on the jont dtrbuton of (X 1,...,X m ). Denote x + := max{0, x}. Note that: B = teady-tate number of backordered component = [X +, depend on the margnal dtrbuton of X only. So t eay to compute. To dentfy B,tmportant to fgure out the correpondng component backlog that contrbute to t. Let B be the teady-tate number of backlogged unt of component that are due to type- demand. We have that: B ( B λ E [ Q ) = 1 k J λ J E [ B ( λ E [ Q ) Q J = 1 k, µ (6) where the 1 k (p) are..d. Bernoull random varable wth parameter p. Inthe cae of unt demand, the number of backorder of type : B { } = max B, where B = B ( λ ) 1 k. (7) λ When demand arrve n batche, however, we mut frt tranlate B nto batche. Let: { } n B Q, := nf n : Q (k) B, (8) where Q (k), k = 1,...,n, are ndependent cope of Q. Then, B = max BQ,. (9) Let c be the unt cot of component. Let w 0be aweghtng factor for the average backorder for product. Itmeaure the relatve mportance of ervce provded to type- order. Denote := ( 1,..., m ) and Z+ m the m- dmenonal non-negatve nteger vector pace. We are ntereted n the followng optmzaton problem: (P) mn w E [B ().t. c c m m C, Z+ m. (10) where C > 0gven contant, repreentng the total nventory budget. In ndutral applcaton, nventory budget often apple to afety tock only. Snce nventory can be dvded nto work-n-proce (WIP) and afety tock, and the WIP part ndependent of the bae-tock level, the contrant n the above problem formulaton content wth practce. (Alo ee Sherbrooke (1992)). We note that, ntead of a fxed budget on the target afety tock level, another commonly een formulaton to mnmze average nventoryholdng and backorder-penalty cot. Solvng th type of problem wll be requre a dfferent et of technque from thoe developed here; ee Lu and Song (2002) for the untdemand cae. A ponted out n the Introducton, the objectve functon n Equaton (10) relate to a delay objectve va Lttle law: Suppoe we want to mnmze: w E [W (), where W denote the delay (repone tme) n fllng type- order, and w the aocated cot. Then, we can wrte: w E[W () = w λ E [B (), hence, we can et w = w /λ n Equaton (10). Obvouly, the objectve functon n (P) very dffcult to evaluate due to the nvolvement of the jont dtrbuton of (X 1,...,X m ) and other ntrcate relaton, uch a the max operaton. In the ret of the paper we develop two approxmate objectve functon whch nvolve only the margnal dtrbuton of X.Wefurther relax the nteger requrement of the decon varable and treat the anon-negatve real value, a they are nvolved n bound and approxmaton. In many applcaton, the are large value; hence, gnorng the ntegralty approprate. The reult two urrogate problem, and we develop effcent oluton approache to olve them. 3. The lower-bound approach We now develop an approxmaton of E [B that ue margnal dtrbuton of X only. One key tep to apply Jenen nequalty to a max operaton, reultng n a lower bound. Therefore, we call th a lower-bound approach. Note that from Equaton (8) we know that B Q, a toppng tme wth repect to {Q (1), Q (2),...,n}, and Q (k) B, B Q, a.. Takng expectaton on both de and applyng Wald dentty on the left-hand de yeld: E [ Q [ Q, [ E B E B. (11)

5 Backorder mnmzaton n multproduct aemble-to-order ytem 767 Now takng expectaton on both de of Equaton (6), and applyng Wald dentty to the ummaton on the rghthand de, we have: E [ B λ E [B E [ Q =. (12) µ Hence, makng ue of Jenen nequalty along wth Equaton (11) and (12), we obtan: [ E [B = E max = max BQ, { λ E [B µ max E [ B Q, E [ B max E [ Q }. (13) We now ue the lower bound n Equaton (??) a a urrogate for the objectve functon n Equaton (??). Defne: b ( ) = E [B ( ) = E [X +. Then, the reultng lower-bound problem can be expreed a: { } (LB) mn w λ b ( ) max, (14) µ.t. c c m m C; Z+ m. Clearly, the objectve functon n (LB) much eaer to evaluate becaue only the margnal dtrbuton are nvolved. Two extreme cae of th problem have been tuded n the lterature. The frt cae that the component demand are ndependent,.e., q = 0f not a ngleton. In th cae the problem reduce to: m mn w b ( ), =1.t. c c m m C: Z+ m. A nce feature of th problem that the objectve functon a eparable convex functon. Sherbrooke (1992) preent agreedy algorthm to olve th problem. The econd extreme cae a ngle-product ATO ytem,.e., ={1,...,m} the only demand type and the BOM fxed. We can alo redefne the unt of each component, o that we obtan a unt demand model. In th cae, the problem become: mn max{b 1 ( 1 ),...,b m ( m )}, (15).t. c c m m C; Z m +. Ung the property that each b ( )decreang and convex, Song and Yao (2002) how that a greedy algorthm can guarantee optmalty. The general cae, however, much more complcated and le tructured becaue we need to deal wth the um of maxmum over overlappng ubet. To ncreae tractablty, n the followng we further relax the nteger requrement n (LB), and treat each a a non-negatve real varable. Let R m + denote the m-dmenonal non-negatve real vector pace. The relaxed problem then: { } (LB ) mn w λ b1 ( 1 ) max, (16) µ.t. c c m m C; R m +. The reultng objectve value a lower bound for (LB ), whch n turn tll a lower bound on the orgnal problem. Although the objectve functon convex, t not dfferentable, o the conventonal tool for convex program are not applcable. In the ret of th ecton we wll dentfy pecal tructure of the problem and how that we can tranform the problem nto one that mnmze the maxmum over non-overlappng et, whch n turn can be converted nto a et of ubproblem wth lnear objectve functon and ordered varable. The prce we pay that the orgnal lnear contrant functon now become a non-lnear decreang functon. Nonethele, we wll how that each of thee ubproblem greedly olvable, o we can obtan the overall optmal oluton effcently. Once we obtan a oluton for (LB ), we round t to the nearet nteger oluton a an approxmate oluton for the orgnal problem The tack tructure A change of varable wll better reveal the tructure of (LB ). Note that b ( ) = E [X + trctly decreang convex n R 1 +. Th mple that t nvere b 1 ( )well defned and alo decreang and convex. Snce lm b ( ) = 0, we denote b 1 (0) =+.Now,for each = 1, 2,...,m, defne: z := b ( ), or = b 1 (µ z ):= h (z ). (17) µ Let z := (z 1,...,z m ). Then, the problem. (LB )equvalent to: (LB ) mn w λ max z (18) z.t. c h (z ) C; z R m +. Suppoe z = (z1,...,z m ) an optmal oluton to Equaton (18). Denote u := max z.obvouly, f then u u.wthout lo of generalty, rename the product type n acendng order of u : u 1 u 2. We now argue that we can aume that all z, 1 have the ame value. Th becaue f there ext an 0 1 uch that: z 0 < u 1 := max zj, j 1 then, reagnng z 0 the value u 1 wll not detroy the optmalty of z : the objectve value wll reman the ame and the contrant wll tll be atfed becaue h 0 ( )adecreang functon.

6 768 Lu et al. Smlarly, note that u 2 := max j 2 z j the econd mallet of the u value. For any 2 \ 1,fz < u 2, then we can ncreae t value to u 2 wthout changng the objectve value or volatng the contrant o to preerve the optmalty of z. Contnung th argument, we can conclude that there ext an optmal oluton z to Equaton (18) that atfe a tack-lke tructure a formalzed below. Suppoe ={ 1,..., n },.e., there are total of n product (or, demand) type. Then we can wrte the et of all component a: I = n l=1 l := {1,...,m}. Conder a partcular order (permutaton) of the n ubet, π = (π 1,π 2,...,π n ). Form n new ubet nductvely a follow: π 1 = π1, π l = πl \( π 1 π l 1 ), l = 2,...,n. (19) Then, clearly, ( π 1,..., π n )apartton of I (wherea ( π1,..., πn )not). Note that f n > m then ome π l wll be empty. Propoton 1. There ext an optmal oluton z to Equaton (18) atfyng the followng tack-lke tructure. Suppoe π = (π 1,π 2,...,π n ) a permutaton uch that u πl := max j πl zj ncreang n l,.e: u π1 u π2 u πn. Let ( π 1,..., π n ) be the partton of I pecfed n Equaton (19). Then, we have: z j = u πj, for all l = 1,...,n. j π l 3.2. The optmzaton algorthm To mplfy notaton, denote, for each permutaton π = (π 1,π 2,...,π n ): λ π l := λ π l, w π l := w π l, uπl := u πl, u π = (u πl,...,u πn ). Furthermore, baed on Propoton 1, denote: g l (z) := π l c h (z). We can then rewrte the optmzaton problem n Equaton (18) a follow: n (LB ) mn mn w π l λ π l u πl, (20) π u π l=1 n ( ).t. g l uπl C, (21) l=1 u π1 u πn, u π R n +. Hence, gven the permutaton π, denote the nner mnmzaton part of the above problem, ncludng the contrant, a (LB π ). We can olve (LB π )forall π; the one that yeld the mallet objectve value then the optmal oluton to the orgnal problem (LB ). For each π, (LB π )equvalent to a problem of mnmzng a eparable convex functon wth a ngle lnear contrant, hence, t greedly olvable va a margnal allocaton algorthm. To decrbe th algorthm, reformulate the problem wth yet another tranformaton of varable: y 1 := u π1 ; y l := u πl u πl 1, l = 2,...,n. Hence, u πl = y 1 + +y l, l = 2,...,n. Alo denote y := (y 1,...,y n ); v l := w π l λ πl + +w π n λ π n, l = 1,...,n. Then, (LB π ) become the followng: n (LB π ) mn v l y l, (22) y l=1 n.t. g l (y 1 + +y l ) C, y R n +. l=1 The margnal allocaton cheme work a follow: Start wth y = 0; hence, the zero objectve value. In each tep, allow an ncreae of >0ofthe objectve value, wth beng aprepecfed contant. Th correpond to allowng y k to ncreae by an amount /v k,for each k = 1,...,n. (Examnng thee vertex pont uffcent becaue g l ( )convex.) Th, n turn, correpond to the decreae of the left-hand de of the contrant by the followng amount: n [g l (y 1 + +y k + +y l ) l=k g l (y 1 + +y k + /v k + +y l ). (Recall, g l ( ) adecreang functon, for all l.) Therefore, elect the k that yeld the larget uch decreae, and update y k to y k + /v k.repeat th procedure untl the left-hand de of the contrant reduced to C Heurtc Although for each π, (LB π )greedly olvable, we tll need to olve n! uch problem. One heurtc to do parwe mprovement: tart from any permutaton, ay, (1,...,n). Compare th wth all other permutaton that reult from nterchangng 1 wth l 1. Th nvolve olvng n margnal allocaton problem lke (LB π )above. Suppoe wappng 1 and π 1 gve the mallet objectve value. Next, fx π 1 at the frt poton, apply the ame parwe nterchange to the remanng n 1 element of the permutaton, and o forth. Th heurtc requre olvng n(n 1)/2 (a oppoed to n!) margnal allocaton problem.

7 Backorder mnmzaton n multproduct aemble-to-order ytem 769 An even mpler heurtc to apply the margnal allocaton cheme that olve (LB π ) drectly to the problem (LB ) n Equaton (20). The detal are a follow: Start wth π = (1,...,n), and et uπl = 0forall l. Suppoe the current oluton (uπl ), correpondng to ome permutaton π = (π l ). For each k = 1,...,n, do: (a) ncreae u πk to u πk + /(w π k λ π k ); rearrange the u value, and let π k denote the reultng permutaton; (b) evaluate the left-hand de of the contrant n Equaton (21). Select k a the one correpondng to the mallet value n (b) above (.e., the one that yeld the larget decreae n the left-hand de of the contrant). Update u πk to u πk + /(w π k λ π k ); and π to π k. Repeat the above untl the left-hand de of the contrant n Equaton (21) reduced to C. Remark 1. Note that wth any gven permutaton the orgnal dcrete lower-bound urrogate problem (LB) a eparable dcrete convex programmng, whch can be ealy olved by a greedy algorthm. So the heurtc algorthm above can alo be appled to olve the problem n (LB). 4. The upper-bound approach We now wtch to ung an upper-bound approach to derve urrogate, whch nclude an upper bound and ome approxmaton derved from upper bound, for the objectve functon n the backorder mnmzaton problem n Equaton (10). The man dea the followng nequalty: for a et of non-negatve varable, x 0, we have max{x } a + (x a) +, (23) whch hold for any a, and can be drectly verfed. Hence, mnmzng the rght-hand de above, wth repect to a, yeld an upper bound for max {x }. Note, however, that when a < 0, the rght-hand de become: a + (x a) x. So, n mnmzng the rght-hand de of Equaton (23), we only need to conder a 0. When (x )avector of non-negatve random varable, the upper bound ha the advantage of nvolvng only the margnal dtrbuton. Hence, applyng th to Equaton (9), we have: B := max { B Q, } a + ( Q, B a ) +, a.., (24) where a a parameter. Therefore, ung the above upper bound a a urrogate objectve, and wrtng a := (a ) (and := ( ) I a before), we have the followng optmzaton problem: E ( B Q, a ) +, (25) mn (a,) w a + w.t. c c m m C, R m +. In the above problem, mnmzng over a not a mportant a mnmzng over,athe former meant to mprove on the upper bound. (In fact, a we hall llutrate below, we could very well forgo the mnmzaton over a by fxng t at a pecfc value.) Hence, we propoe to ue a common (calar) a for all product type, and then optmze a. Th way, the problem become: (UB) mn (a,) a w + w.t. c c m m C, 0. E ( B Q, a ) +,(26) For fxed a, (UB) a eparable convex programmng, whch greedly olvable. We can then conduct a lne earch on a. Alternatvely, electng a = E(B Q, ), we have: B { ( Q,)} max E B + E [ B Q, E ( B Q, ) +, a.. Takng expectaton on both de yeld: E [B { ( Q,)} max E B + E [ B Q, E ( B Q, ) +. (27) Note that from Equaton (8) we have: (B Q, 1) + Q (k), a.., B where Q (k) are ndependent cope of Q and 0 Q (k) = 0. Applyng Wald dentty yeld: E [ B E [ Q E [ B Q, 1 +. (28) It reaonable to ue B /E [Q toapproxmate B Q,. Th relatonhp exact for the unt demand model. Conequently, we can ue E [B /E [Q toapproxmate E [B Q,. Apply thee approxmaton to Equaton (27), we obtan: { [ } E B E [B max E [ Q + E [ B E ( ) B + E [ Q. (29) We now go one tep further by replacng the ummaton over n the rght-hand de of Equaton (29) wth the maxmum over, reultng n: { [ E B E [B max E [ Q + E [ B E ( ) B + } E [ Q { λ = max E [ B E [ B + E ( B ) + } µ E [ Q. (30)

8 770 Lu et al. Ung the above approxmaton n the objectve of (P) yeld the followng urrogate problem: { λ (APP) mn w max E [B ( ) µ + E [ B ( ) E B ( ) + } E [ Q, (31).t. c c m m C, 0 R m +. Note that although the orgnal dea wa to ue an upper bound to urrogate the orgnal objectve functon, we took two tep of approxmaton to reve the upper-bound functon to acheve Equaton (30). Therefore, the objectve functon n (APP) no longer an upper bound. In the ret of th ecton we hall how that (APP) exhbt the ame tructure a n (LB ). The followng propoton tell u that E [B ( ) EB ( ) + decreang n. So, the objectve functon n (APP) ha the ame property, and therefore t nvere functon unquely defned. Propoton 2. For any convex functon f (x), E f [B E(B ) decreang n.inpartcular, Var(B ) decreang n, for each and each ; and Var(B ) decreang n, for all. Proof. We frt how that Var(B )decreang n.recall, B = [X +. Hence, t uffce to how that: E [(X ) + 2 E [(X 1) + 2 [E(X ) + 2 [E(X 1) + 2. (32) We have, (X ) + (X 1) + = 1(X + 1), (X ) + + (X 1) + = (2X 2 + 1)1(X + 1). (Note that we are treatng the X, a well a the anteger valued.) Hence: E{[(X ) + 2 [(X 1) + 2 } = E [2X 2 + 1)1(X + 1), wherea [E(X ) + 2 [E(X 1) + 2 = [E(X ) + + E(X 1) + [E (X ) + E(X 1) +, = E [(2X 2 + 1)1(X + 1)P (X + 1). Hence, the nequalty n Equaton (32) hold. Next, wrte B,n a B condtonng upon X = n,for n = 0, 1,...From Equaton (6), nce X ndependent of the Bernoull varable nvolved, t uffce to how that E f [B,n E(B,n ) ncreang n n. Denote M n := B,n E(B,n ). From Equaton (6), clearly, {M n } formamartngale (Ro, 1996). Hence, the convexty of f mple that {f (M n )} a ubmartngale, whch, n turn, mple that E f (M n )ncreang n n, and hence decreang n,whch what dered. Lettng f (x) = x 2,wehaveE(Mn 2) = Var(B,n ). Note that we can wrte: Var ( B ) [ ( ) [ ( ) = E Var B X + Var E B X. (33) The frt term on the rght-hand de above decreang n, nce Var(B,n )ncreang n n gven X = n a argued above. Followng Equaton (6), the econd term equal to (λ E [Q /λ )Var(B ); hence, t alo decreang n,fromwhat ha already been hown for Var(B ). Therefore, we can conclude that Var(B )decreang n. Fnally, oberve from Equaton (6) that B ( )/ (λ E [Q ) and [X + 1 k (1/µ )have the ame frt two moment and the latter ndependent of.wereplace the former wth the latter a an approxmaton. Th lead to the followng problem: { } E (APP) mn w λ [B ( ) max + r( ), (34) µ where r( ) = E.t. c c m m C, R m +. [ [X = + ( ) 1 1 k E [X + +. µ Followng mlar argument n the proof of Propoton 2, we can how that r( )alo a decreang functon of. Therefore, (APP ) ha the ame tructure a the lowerbound problem n the lat ecton. So, the ame optmzaton and heurtc algorthm can be appled here. 5. Numercal example In th ecton, we preent reult from a et of numercal experment, n whch the heurtc algorthm derved n the prevou ecton are teted for ther effcency n fndng the (near-)optmal oluton to the backorder mnmzaton problem. A mentoned before, what complcate the multproduct problem the number of overlappng ubet that conttute dfferent type of product, not the number of component. For ntance, f one et of product hare no common component wth another et of product, then we can mply decompoe the orgnal ytem nto two eparate ytem, each of whch cover one et of product. An extreme cae that each product cont of only a ngle component, reultng n a eparable convex problem whch greedly olvable (ee Sherbrooke (1992)). For th purpoe, n our experment we chooe example that are complex n the ene that there are many common component hared by dfferent product. We conder an ATO ytem wth x component and the followng x product type: {2, 5}, {3, 5}, {1, 2, 5}, {1, 3, 6}, {1, 3, 4, 5}, {1, 3, 4, 6}. µ

9 Backorder mnmzaton n multproduct aemble-to-order ytem 771 The overall demand arrval rate λ = 4, whch plt nto the x product type a follow: q 25 = 0.10, q 35 = 0.40, q 125 = 0.15, q 136 = 0.10, q 1345 = 0.20, q 1346 = The component leadtme follow ndependent exponental dtrbuton wth mean: (l ) =1 6 = (1, 1, 1, 1, 2, 2). We conder two et of cot data: c (1) = (1, 1, 1, 1, 1, 1) and c (2) = (2, 2, 3, 2, 1, 1), wth repectve weght w (1) = (1, 1, 1, 1, 1, 1) and w (2) = (1, 1, 1.2, 1.5, 1.2, 1.5), and varou nventory budget level: C = 20, 24, 32 for c (1) ; and C = 30, 40, 50 for c (2). In Table 1 and 2, we ummarze the oluton obtaned by: () an exhautve earch va mulaton; () the lowerbound approach along wth the heurtc algorthm n Secton 3.3; () the upper-bound approach baed on (UB); and (v) the approxmaton approach baed on (APP ). For each approach, we report the optmal bae-tock level, the correpondng objectve value, and the reultng weghted average backorder. Note that whle the lat two meaure are equal for the mulaton-baed optmal oluton, they are dfferent for the other three approache. For ntance, the objectve value correpondng to () necearly a lower bound of the objectve value of (); wherea the weghted average backorder correpondng to () obtaned from ung the optmal bae-tock level derved from the lower-bound approach (to evaluate analytcally the orgnal objectve). We alo report the percentage lo of optmalty of each approxmate approach. In all of our numercal example, the leadtme follow exponental dtrbuton; the mpact of dfferent dtrbutonal aumpton upon the performance of thee ATO ytem can be found n [11. It alo ponted out n [11 that the dfference caued by dtrbutonal aumpton not gnfcant epecally for hgh performance ytem. Therefore, we beleve that thee aumpton wll not affect the accuracy of our approxmaton. In Table 3 and 4, we repeat the ame example wth λ = 8, and modfed budget level: C = 30, 36, 45 for c (1) ; and C = 40, 50, 60 for c (2). The reult ndcate that oluton from the urrogate problem are very cloe to the true optmal oluton n all cae. Overall, the approxmaton approach (APP )appear to be the mot effectve method. We now ue thee example to hed ome lght on how the optmal component nventory level dffer from a tandard ngle-tem approach ued n practce. Recall that the ngletem nventory theory ugget that we hould tock enough of each component to cover the leadtme demand plu ome afety tock to protect agant varablty (of both demand and leadtme). Accordng to Equaton (1), (3) and (4), Table 1. Reult obtaned for the ytem contanng the x product type q 25 = 0.10, q 35 = 0.40, q 125 = 0.15, q 136 = 0.10, q 1345 = 0.20 and q 1346 = Alo, L = (1,1,1,1,2,2) and the overall arrval rate equal to four Coeffcent Weght Budget Algorthm Objectve Ave. backorder Rel. err. (%) (1,1,1,1,1,1) (1,1,1,1,1,1) C = 20 optmal lower bound upper bound approxmaton C = 24 optmal lower bound upper bound approxmaton C = 32 optmal lower bound upper bound approxmaton (2,2,3,2,1,1) (1,1,1,1,1,1) C = 30 optmal lower bound upper bound approxmaton C = 40 optmal lower bound upper bound approxmaton C = 50 optmal lower bound upper bound approxmaton

10 772 Lu et al. Table 2. Reult obtaned for the ytem contanng x product type q 25 = 0.10, q 35 = 0.40, q 125 = 0.15, q 136 = 0.10, q 1345 = 0.20 and q 1346 = Alo, L = (1,1,1,1,2,2) and the overall arrval rate equal to four Coeffcent Weght Budget Algorthm Objectve Ave. backorder Rel. err. (%) (1,1,1,1,1,1) (1,1,1.2,1.5,1.2,1.5) C = 20 optmal lower bound upper bound approxmaton C = 24 optmal lower bound upper bound approxmaton C = 32 optmal lower bound upper bound approxmaton (2,2,3,2,1,1) (1,1,1.2,1.5,1.2,1.5) C = 30 optmal lower bound upper bound approxmaton C = 40 optmal lower bound upper bound approxmaton C = 50 optmal lower bound upper bound approxmaton Table 3. Reult obtaned for the ytem contanng x product type q 25 = 0.10, q 35 = 0.40, q 125 = 0.15, q 136 = 0.10, q 1345 = 0.20 and q 1346 = Alo, L = (1, 1, 1, 1, 2, 2) and the overall arrval rate equal to eght Coeffcent Weght Budget Algorthm Objectve Ave. backorder Rel. err. (%) (1,1,1,1,1,1) (1,1,1,1,1,1) C = 30 optmal lower bound upper bound approxmaton C = 36 optmal lower bound upper bound approxmaton C = 45 optmal lower bound upper bound approxmaton (2,2,3,2,1,1) (1,1,1,1,1,1) C = 40 optmal lower bound upper bound approxmaton C = 50 optmal lower bound upper bound approxmaton C = 60 optmal lower bound upper bound approxmaton

11 Backorder mnmzaton n multproduct aemble-to-order ytem 773 Table 4. Reult obtaned for the ytem contanng x product type q 25 = 0.10, q 35 = 0.40, q 125 = 0.15, q 136 = 0.10, q 1345 = 0.20, and q 1346 = Alo, L = (1,1,1,1,2,2) and the overall arrval rate equal to eght Coeffcent Weght Budget Algorthm Objectve Ave. backorder Rel. err. (%) (1,1,1,1,1,1) (1,1,1.2,1.5,1.2,1.5) C = 30 optmal lower bound upper bound approxmaton C = 36 optmal lower bound upper bound approxmaton C = 45 optmal lower bound upper bound approxmaton (2,2,3,2,1,1) (1,1,1.2,1.5,1.2,1.5) C = 40 optmal lower bound upper bound approxmaton C = 50 optmal lower bound upper bound approxmaton C = 60 optmal lower bound upper bound approxmaton Table 5. Optmal and approxmate afety factor Arrval rate Coeffcent Weght Budget k 1 k 1 k 2 k 2 k 3 k 3 k 4 k 4 k 5 k 5 k 6 k 6 4 (1,1,1,1,1,1) (1,1,1,1,1,1) (2,2,3,2,1,1) (1,1,1,1,1,1) (1,1,1,1,1,1) (1,1,1.2,1.5,1.2,1.5) (2,2,3,2,1,1) (1,1,1.2,1.5,1.2,1.5) (1,1,1,1,1,1) (1,1,1,1,1,1) (2,2,3,2,1,1) (1,1,1,1,1,1) (1,1,1,1,1,1) (1,1,1.2,1.5,1.2,1.5) (2,2,3,2,1,1) (1,1,1.2,1.5,1.2,1.5)

12 774 Lu et al. we have: = θ + k σ, (35) where θ the average leadtme demand for component ; σ, the correpondng tandard devaton; and k, the afety factor, a potve parameter ndcatve of the level of protecton agant uncertanty or varablty. When there are multple tem (component), one mut addre how to chooe k for each. What often followed n practce the o-called equal fractle rule: havng the ame k value acro all component. Th equvalent to projectng the ame level of fll rate acro. Should th be the cae here? To anwer th queton, n Table 5 we report the optmal afety factor aocated wth the optmal bae-tock level n the prevou four table. From thee afety factor, we can derve the component fll rate from a normal dtrbuton table. For ntance, k = 1.00 correpond to a fll rate of (1) = 84.13%, where (x) denote the dtrbuton functon of the tandard normal varate; k = 2 correpond to a fll rate of (2) = 97.72%, and o forth. A can be oberved from the table, the optmal afety factor vary wdely among component. In ome cae, when the budget level relatvely low, the afety factor of certan component are even negatve. Thee obervaton ugget that the equal fractle rule n general not effectve n ATO ytem. Next to each optmal afety factor k,wealo report the approxmate afety factor k derved from the approxmate (APP ) oluton. Clearly, thee example ndcate that ung the approxmate approach we can come up wth cloe-to-optmal afety factor. To ummarze, among the three urrogate problem and algorthm developed here, both (LB) and (UB) have a computatonal advantage and can provde a good predcton of the bac trend of the ytem behavor under varou change. On the other hand, (APP ), wth more computatonal effort than thoe of (LB) and (UB), provde a good approxmaton to both the optmal oluton and the objectve value. (In cae where (APP ) doe not produce the bet reult, the relatve error of the optmal objectve are below 3%.) Reference Agrawal, M. and Cohen, M. (2001) Optmal materal control and performance evaluaton n an aembly envronment wth component commonalty. Naval Reearch Logtc, 48, Cheng, F., Ettl, M., Ln, G.Y. and Yao, D.D. (2002) Inventory-ervce optmzaton n confgure-to-order ytem. Manufacturng Servce Operaton Management, 4, Cheung,.L. and Hauman, W. (1995) Multple falure n a mult-tem pare nventory model. IIE Tranacton, 27, Glaerman, P. and Wang, Y. (1998) Leadtme-nventory tradeoff n aemble-to-order ytem. Operaton Reearch, 46, Gallen, J. and Wen, L. (2001) A mple and effectve component procurement polcy for tochatc aembly ytem. Queueng Sytem, 38, Hauman, W.H., Lee, H.L. and Zhang, A.X. (1998) Order repone tme relablty n a mult-tem nventory ytem. European Journal of Operatonal Reearch, 109, ouvel, P., arabat, S. and Yu, G. (2001) A mn-max-um reource allocaton problem and t applcaton. Operaton Reearch, 49, Lu, Y. and Song, J.S. (2005) Order-baed cot optmzaton n aembleto-order ytem. Operaton Reearch, 53, Lu, Y., Song, J.S. and Yao, D. D. (2003) Order fll rate, leadtme varablty, and advance demand nformaton n an aemble-to-order ytem. Operaton Reearch, 51, Pne, B.J. (1993) Ma Cutomzaton: The New Fronter n Bune Competton, Harvard Bune School Pre, Cambrdge, MA. Ro, S.M. (1996) Stochatc Procee, 2nd edn., Wley, New York, NY. Shanthkumar, J.G. and Yao, D.D. (1991b) Bvarate characterzaton of ome tochatc order relaton. Advance n Appled Probablty, 23, Sherbrooke, C.C. (1992) Optmal Inventory Modelng of Sytem, Wley, New York, NY. Song, J.S. and Yao, D.D. (2002). Performance analy and optmzaton n aemble-to-order ytem wth random leadtme. Operaton Reearch, 50, Song, J.S. and Zpkn, P. (2003). Supply chan operaton: aembleto-order ytem, n Supply Chan Management, De ok, T. and Grave, S. (ed.), North-Holland, Amterdam, The Netherland, Ch. 11. Wang, Y. (1999) Near-optmal bae-tock polce n aemble-to-order ytem under ervce level requrement. Preprnt, MIT Sloan School, MIT, Cambrdge, MA. Wolff, R. (1989) Stochatc Modelng and the Theory of Queue,Prentce Hall, Englewood Clff, NJ. Bographe Yngdong Lu a reearch taff member at the IBM T.J. Waton Reearch Center. H reearch nteret nclude appled probablty and tochatc model. Jng-Sheng (Jeannette) Song a Profeor at the Fuqua School of Bune, Duke Unverty. She obtaned her Ph.D. from Columba Unverty. Prevouly, he ha alo held faculty poton at the Unverty of Calforna at Irvne and Columba Unverty. She tude the degn and management of producton, nventory, and logtc ytem and coordnaton mechanm n upply chan. Her reearch ha been upported by the Natonal Scence Foundaton and ha appeared n everal leadng academc journal. She erve on the Edtoral Board of Management Scence, Operaton Reearch, Manufacturng & Servce Operaton Management, IIE Tranacton, and Naval Reearch Logtc. Davd Yao receved h Ph.D. degree from the Unverty of Toronto n 1983, and tarted h academc career at Columba Unverty, where he became a full profeor n He ha performed extenve reearch and conultng work n tochatc network, emconductor manufacturng, and upply chan management. He the author/co-author of over 160 refereed publcaton, three book and fve edted volume. He a recpent of numerou honor and award, an IEEE Fellow, and a holder of four US patent n manufacturng operaton and upply-chan logtc. He the Stochatc Model Area Edtor of Operaton Reearch, and Edtor-n-Chef of Foundaton and Trend n Stochatc Sytem. Contrbuted by the Supply Chan/Producton-Inventory Sytem Department

Additional File 1 - Detailed explanation of the expression level CPD

Additional File 1 - Detailed explanation of the expression level CPD Addtonal Fle - Detaled explanaton of the expreon level CPD A mentoned n the man text, the man CPD for the uterng model cont of two ndvdual factor: P( level gen P( level gen P ( level gen 2 (.).. CPD factor

More information

Backorder Minimization in Multiproduct Assemble-to-Order Systems

Backorder Minimization in Multiproduct Assemble-to-Order Systems Backorder Mnmzaton n Multproduct Assemble-to-Order Systems Yngdong Lu IBM T.J. Watson Research Center Yorktown Heghts, NY 10598 Tel. (914) 945-3738, e-mal: yngdong@us.bm.com Jng-Sheng Song The Fuqua School

More information

Specification -- Assumptions of the Simple Classical Linear Regression Model (CLRM) 1. Introduction

Specification -- Assumptions of the Simple Classical Linear Regression Model (CLRM) 1. Introduction ECONOMICS 35* -- NOTE ECON 35* -- NOTE Specfcaton -- Aumpton of the Smple Clacal Lnear Regreon Model (CLRM). Introducton CLRM tand for the Clacal Lnear Regreon Model. The CLRM alo known a the tandard lnear

More information

Chapter 11. Supplemental Text Material. The method of steepest ascent can be derived as follows. Suppose that we have fit a firstorder

Chapter 11. Supplemental Text Material. The method of steepest ascent can be derived as follows. Suppose that we have fit a firstorder S-. The Method of Steepet cent Chapter. Supplemental Text Materal The method of teepet acent can be derved a follow. Suppoe that we have ft a frtorder model y = β + β x and we wh to ue th model to determne

More information

Improvements on Waring s Problem

Improvements on Waring s Problem Improvement on Warng Problem L An-Png Bejng, PR Chna apl@nacom Abtract By a new recurve algorthm for the auxlary equaton, n th paper, we wll gve ome mprovement for Warng problem Keyword: Warng Problem,

More information

Chapter 6 The Effect of the GPS Systematic Errors on Deformation Parameters

Chapter 6 The Effect of the GPS Systematic Errors on Deformation Parameters Chapter 6 The Effect of the GPS Sytematc Error on Deformaton Parameter 6.. General Beutler et al., (988) dd the frt comprehenve tudy on the GPS ytematc error. Baed on a geometrc approach and aumng a unform

More information

Estimation of Finite Population Total under PPS Sampling in Presence of Extra Auxiliary Information

Estimation of Finite Population Total under PPS Sampling in Presence of Extra Auxiliary Information Internatonal Journal of Stattc and Analy. ISSN 2248-9959 Volume 6, Number 1 (2016), pp. 9-16 Reearch Inda Publcaton http://www.rpublcaton.com Etmaton of Fnte Populaton Total under PPS Samplng n Preence

More information

Harmonic oscillator approximation

Harmonic oscillator approximation armonc ocllator approxmaton armonc ocllator approxmaton Euaton to be olved We are fndng a mnmum of the functon under the retrcton where W P, P,..., P, Q, Q,..., Q P, P,..., P, Q, Q,..., Q lnwgner functon

More information

Solution Methods for Time-indexed MIP Models for Chemical Production Scheduling

Solution Methods for Time-indexed MIP Models for Chemical Production Scheduling Ian Davd Lockhart Bogle and Mchael Farweather (Edtor), Proceedng of the 22nd European Sympoum on Computer Aded Proce Engneerng, 17-2 June 212, London. 212 Elever B.V. All rght reerved. Soluton Method for

More information

Team. Outline. Statistics and Art: Sampling, Response Error, Mixed Models, Missing Data, and Inference

Team. Outline. Statistics and Art: Sampling, Response Error, Mixed Models, Missing Data, and Inference Team Stattc and Art: Samplng, Repone Error, Mxed Model, Mng Data, and nference Ed Stanek Unverty of Maachuett- Amhert, USA 9/5/8 9/5/8 Outlne. Example: Doe-repone Model n Toxcology. ow to Predct Realzed

More information

MULTIPLE REGRESSION ANALYSIS For the Case of Two Regressors

MULTIPLE REGRESSION ANALYSIS For the Case of Two Regressors MULTIPLE REGRESSION ANALYSIS For the Cae of Two Regreor In the followng note, leat-quare etmaton developed for multple regreon problem wth two eplanator varable, here called regreor (uch a n the Fat Food

More information

Pythagorean triples. Leen Noordzij.

Pythagorean triples. Leen Noordzij. Pythagorean trple. Leen Noordz Dr.l.noordz@leennoordz.nl www.leennoordz.me Content A Roadmap for generatng Pythagorean Trple.... Pythagorean Trple.... 3 Dcuon Concluon.... 5 A Roadmap for generatng Pythagorean

More information

Order Full Rate, Leadtime Variability, and Advance Demand Information in an Assemble- To-Order System

Order Full Rate, Leadtime Variability, and Advance Demand Information in an Assemble- To-Order System Order Full Rate, Leadtme Varablty, and Advance Demand Informaton n an Assemble- To-Order System by Lu, Song, and Yao (2002) Presented by Png Xu Ths summary presentaton s based on: Lu, Yngdong, and Jng-Sheng

More information

Introduction to Interfacial Segregation. Xiaozhe Zhang 10/02/2015

Introduction to Interfacial Segregation. Xiaozhe Zhang 10/02/2015 Introducton to Interfacal Segregaton Xaozhe Zhang 10/02/2015 Interfacal egregaton Segregaton n materal refer to the enrchment of a materal conttuent at a free urface or an nternal nterface of a materal.

More information

A METHOD TO REPRESENT THE SEMANTIC DESCRIPTION OF A WEB SERVICE BASED ON COMPLEXITY FUNCTIONS

A METHOD TO REPRESENT THE SEMANTIC DESCRIPTION OF A WEB SERVICE BASED ON COMPLEXITY FUNCTIONS UPB Sc Bull, Sere A, Vol 77, I, 5 ISSN 3-77 A METHOD TO REPRESENT THE SEMANTIC DESCRIPTION OF A WEB SERVICE BASED ON COMPLEXITY FUNCTIONS Andre-Hora MOGOS, Adna Magda FLOREA Semantc web ervce repreent

More information

Two Approaches to Proving. Goldbach s Conjecture

Two Approaches to Proving. Goldbach s Conjecture Two Approache to Provng Goldbach Conecture By Bernard Farley Adved By Charle Parry May 3 rd 5 A Bref Introducton to Goldbach Conecture In 74 Goldbach made h mot famou contrbuton n mathematc wth the conecture

More information

This appendix presents the derivations and proofs omitted from the main text.

This appendix presents the derivations and proofs omitted from the main text. Onlne Appendx A Appendx: Omtted Dervaton and Proof Th appendx preent the dervaton and proof omtted from the man text A Omtted dervaton n Secton Mot of the analy provded n the man text Here, we formally

More information

Start Point and Trajectory Analysis for the Minimal Time System Design Algorithm

Start Point and Trajectory Analysis for the Minimal Time System Design Algorithm Start Pont and Trajectory Analy for the Mnmal Tme Sytem Degn Algorthm ALEXANDER ZEMLIAK, PEDRO MIRANDA Department of Phyc and Mathematc Puebla Autonomou Unverty Av San Claudo /n, Puebla, 757 MEXICO Abtract:

More information

Root Locus Techniques

Root Locus Techniques Root Locu Technque ELEC 32 Cloed-Loop Control The control nput u t ynthezed baed on the a pror knowledge of the ytem plant, the reference nput r t, and the error gnal, e t The control ytem meaure the output,

More information

Scattering of two identical particles in the center-of. of-mass frame. (b)

Scattering of two identical particles in the center-of. of-mass frame. (b) Lecture # November 5 Scatterng of two dentcal partcle Relatvtc Quantum Mechanc: The Klen-Gordon equaton Interpretaton of the Klen-Gordon equaton The Drac equaton Drac repreentaton for the matrce α and

More information

m = 4 n = 9 W 1 N 1 x 1 R D 4 s x i

m = 4 n = 9 W 1 N 1 x 1 R D 4 s x i GREEDY WIRE-SIZING IS LINEAR TIME Chr C. N. Chu D. F. Wong cnchu@c.utexa.edu wong@c.utexa.edu Department of Computer Scence, Unverty of Texa at Autn, Autn, T 787. ABSTRACT In nterconnect optmzaton by wre-zng,

More information

The multivariate Gaussian probability density function for random vector X (X 1,,X ) T. diagonal term of, denoted

The multivariate Gaussian probability density function for random vector X (X 1,,X ) T. diagonal term of, denoted Appendx Proof of heorem he multvarate Gauan probablty denty functon for random vector X (X,,X ) px exp / / x x mean and varance equal to the th dagonal term of, denoted he margnal dtrbuton of X Gauan wth

More information

Variable Structure Control ~ Basics

Variable Structure Control ~ Basics Varable Structure Control ~ Bac Harry G. Kwatny Department of Mechancal Engneerng & Mechanc Drexel Unverty Outlne A prelmnary example VS ytem, ldng mode, reachng Bac of dcontnuou ytem Example: underea

More information

Small signal analysis

Small signal analysis Small gnal analy. ntroducton Let u conder the crcut hown n Fg., where the nonlnear retor decrbed by the equaton g v havng graphcal repreentaton hown n Fg.. ( G (t G v(t v Fg. Fg. a D current ource wherea

More information

Synchronization Protocols. Task Allocation Bin-Packing Heuristics: First-Fit Subtasks assigned in arbitrary order To allocate a new subtask T i,j

Synchronization Protocols. Task Allocation Bin-Packing Heuristics: First-Fit Subtasks assigned in arbitrary order To allocate a new subtask T i,j End-to-End Schedulng Framework 1. Tak allocaton: bnd tak to proceor 2. Synchronzaton protocol: enforce precedence contrant 3. Subdeadlne agnment 4. Schedulablty analy Tak Allocaton Bn-Packng eurtc: Frt-Ft

More information

Designing Service Competitions among Heterogeneous Suppliers

Designing Service Competitions among Heterogeneous Suppliers Degnng Servce Competton among Heterogeneou Suppler Ehan Elah and Saf Benaafar Graduate Program n Indutral & Sytem Engneerng Department of Mechancal Engneerng Unverty of Mnneota, Mnneapol, M 55455 elah@me.umn.edu

More information

Generalized Linear Methods

Generalized Linear Methods Generalzed Lnear Methods 1 Introducton In the Ensemble Methods the general dea s that usng a combnaton of several weak learner one could make a better learner. More formally, assume that we have a set

More information

Method Of Fundamental Solutions For Modeling Electromagnetic Wave Scattering Problems

Method Of Fundamental Solutions For Modeling Electromagnetic Wave Scattering Problems Internatonal Workhop on MehFree Method 003 1 Method Of Fundamental Soluton For Modelng lectromagnetc Wave Scatterng Problem Der-Lang Young (1) and Jhh-We Ruan (1) Abtract: In th paper we attempt to contruct

More information

Information Acquisition in Global Games of Regime Change (Online Appendix)

Information Acquisition in Global Games of Regime Change (Online Appendix) Informaton Acquton n Global Game of Regme Change (Onlne Appendx) Mchal Szkup and Iabel Trevno Augut 4, 05 Introducton Th appendx contan the proof of all the ntermedate reult that have been omtted from

More information

STOCHASTIC BEHAVIOUR OF COMMUNICATION SUBSYSTEM OF COMMUNICATION SATELLITE

STOCHASTIC BEHAVIOUR OF COMMUNICATION SUBSYSTEM OF COMMUNICATION SATELLITE IJS 4 () July Sharma & al ehavour of Subytem of ommuncaton Satellte SOHSI HVIOU O OMMUNIION SUSYSM O OMMUNIION SLLI SK Mttal eepankar Sharma & Neelam Sharma 3 S he author n th paper have dcued the tochatc

More information

Lecture 10 Support Vector Machines II

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

More information

Discrete Simultaneous Perturbation Stochastic Approximation on Loss Function with Noisy Measurements

Discrete Simultaneous Perturbation Stochastic Approximation on Loss Function with Noisy Measurements 0 Amercan Control Conference on O'Farrell Street San Francco CA USA June 9 - July 0 0 Dcrete Smultaneou Perturbaton Stochatc Approxmaton on Lo Functon wth Noy Meaurement Q Wang and Jame C Spall Abtract

More information

Statistical Properties of the OLS Coefficient Estimators. 1. Introduction

Statistical Properties of the OLS Coefficient Estimators. 1. Introduction ECOOMICS 35* -- OTE 4 ECO 35* -- OTE 4 Stattcal Properte of the OLS Coeffcent Etmator Introducton We derved n ote the OLS (Ordnary Leat Square etmator ˆβ j (j, of the regreon coeffcent βj (j, n the mple

More information

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS Avalable onlne at http://sck.org J. Math. Comput. Sc. 3 (3), No., 6-3 ISSN: 97-537 COMPARISON OF SOME RELIABILITY CHARACTERISTICS BETWEEN REDUNDANT SYSTEMS REQUIRING SUPPORTING UNITS FOR THEIR OPERATIONS

More information

Assortment Optimization under MNL

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

More information

Improvements on Waring s Problem

Improvements on Waring s Problem Imrovement on Warng Problem L An-Png Bejng 85, PR Chna al@nacom Abtract By a new recurve algorthm for the auxlary equaton, n th aer, we wll gve ome mrovement for Warng roblem Keyword: Warng Problem, Hardy-Lttlewood

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

AP Statistics Ch 3 Examining Relationships

AP Statistics Ch 3 Examining Relationships Introducton To tud relatonhp between varable, we mut meaure the varable on the ame group of ndvdual. If we thnk a varable ma eplan or even caue change n another varable, then the eplanator varable and

More information

Resonant FCS Predictive Control of Power Converter in Stationary Reference Frame

Resonant FCS Predictive Control of Power Converter in Stationary Reference Frame Preprnt of the 9th World Congre The Internatonal Federaton of Automatc Control Cape Town, South Afrca. Augut -9, Reonant FCS Predctve Control of Power Converter n Statonary Reference Frame Lupng Wang K

More information

Foresighted Resource Reciprocation Strategies in P2P Networks

Foresighted Resource Reciprocation Strategies in P2P Networks Foreghted Reource Recprocaton Stratege n PP Networ Hyunggon Par and Mhaela van der Schaar Electrcal Engneerng Department Unverty of Calforna Lo Angele (UCLA) Emal: {hgpar mhaela@ee.ucla.edu Abtract We

More information

APPROXIMATE FUZZY REASONING BASED ON INTERPOLATION IN THE VAGUE ENVIRONMENT OF THE FUZZY RULEBASE AS A PRACTICAL ALTERNATIVE OF THE CLASSICAL CRI

APPROXIMATE FUZZY REASONING BASED ON INTERPOLATION IN THE VAGUE ENVIRONMENT OF THE FUZZY RULEBASE AS A PRACTICAL ALTERNATIVE OF THE CLASSICAL CRI Kovác, Sz., Kóczy, L.T.: Approxmate Fuzzy Reaonng Baed on Interpolaton n the Vague Envronment of the Fuzzy Rulebae a a Practcal Alternatve of the Clacal CRI, Proceedng of the 7 th Internatonal Fuzzy Sytem

More information

728. Mechanical and electrical elements in reduction of vibrations

728. Mechanical and electrical elements in reduction of vibrations 78. Mechancal and electrcal element n reducton of vbraton Katarzyna BIAŁAS The Slean Unverty of Technology, Faculty of Mechancal Engneerng Inttute of Engneerng Procee Automaton and Integrated Manufacturng

More information

ENTROPY BOUNDS USING ARITHMETIC- GEOMETRIC-HARMONIC MEAN INEQUALITY. Guru Nanak Dev University Amritsar, , INDIA

ENTROPY BOUNDS USING ARITHMETIC- GEOMETRIC-HARMONIC MEAN INEQUALITY. Guru Nanak Dev University Amritsar, , INDIA Internatonal Journal of Pure and Appled Mathematc Volume 89 No. 5 2013, 719-730 ISSN: 1311-8080 prnted veron; ISSN: 1314-3395 on-lne veron url: http://.jpam.eu do: http://dx.do.org/10.12732/jpam.v895.8

More information

Suggested solutions for the exam in SF2863 Systems Engineering. June 12,

Suggested solutions for the exam in SF2863 Systems Engineering. June 12, Suggested solutons for the exam n SF2863 Systems Engneerng. June 12, 2012 14.00 19.00 Examner: Per Enqvst, phone: 790 62 98 1. We can thnk of the farm as a Jackson network. The strawberry feld s modelled

More information

Alpha Risk of Taguchi Method with L 18 Array for NTB Type QCH by Simulation

Alpha Risk of Taguchi Method with L 18 Array for NTB Type QCH by Simulation Proceedng of the World Congre on Engneerng 00 Vol II WCE 00, July -, 00, London, U.K. Alpha Rk of Taguch Method wth L Array for NTB Type QCH by Smulaton A. Al-Refae and M.H. L Abtract Taguch method a wdely

More information

Batch RL Via Least Squares Policy Iteration

Batch RL Via Least Squares Policy Iteration Batch RL Va Leat Square Polcy Iteraton Alan Fern * Baed n part on lde by Ronald Parr Overvew Motvaton LSPI Dervaton from LSTD Expermental reult Onlne veru Batch RL Onlne RL: ntegrate data collecton and

More information

Optimal inference of sameness Supporting information

Optimal inference of sameness Supporting information Optmal nference of amene Supportng nformaton Content Decon rule of the optmal oberver.... Unequal relablte.... Equal relablte... 5 Repone probablte of the optmal oberver... 6. Equal relablte... 6. Unequal

More information

Analysis of Discrete Time Queues (Section 4.6)

Analysis of Discrete Time Queues (Section 4.6) Analyss of Dscrete Tme Queues (Secton 4.6) Copyrght 2002, Sanjay K. Bose Tme axs dvded nto slots slot slot boundares Arrvals can only occur at slot boundares Servce to a job can only start at a slot boundary

More information

and decompose in cycles of length two

and decompose in cycles of length two Permutaton of Proceedng of the Natona Conference On Undergraduate Reearch (NCUR) 006 Domncan Unverty of Caforna San Rafae, Caforna Apr - 4, 007 that are gven by bnoma and decompoe n cyce of ength two Yeena

More information

OPTIMAL COMPUTING BUDGET ALLOCATION FOR MULTI-OBJECTIVE SIMULATION MODELS. David Goldsman

OPTIMAL COMPUTING BUDGET ALLOCATION FOR MULTI-OBJECTIVE SIMULATION MODELS. David Goldsman Proceedng of the 004 Wnter Smulaton Conference R.G. Ingall, M. D. Roett, J. S. Smth, and B. A. Peter, ed. OPTIMAL COMPUTING BUDGET ALLOCATION FOR MULTI-OBJECTIVE SIMULATION MODELS Loo Hay Lee Ek Peng Chew

More information

The Price of Anarchy in a Network Pricing Game

The Price of Anarchy in a Network Pricing Game The Prce of Anarchy n a Network Prcng Game John Muaccho and Shuang Wu Abtract We analyze a game theoretc model of competng network ervce provder that trategcally prce ther ervce n the preence of elatc

More information

Energy Saving for Automatic Train Control in. Moving Block Signaling System

Energy Saving for Automatic Train Control in. Moving Block Signaling System Energy Savng for Automatc Tran Control n Movng Block Sgnalng Sytem Qng Gu, Xao-Yun Lu and Tao Tang Abtract Wth rapd development of the ralway traffc, the movng block gnalng ytem (MBS) method ha become

More information

Difference Equations

Difference Equations Dfference Equatons c Jan Vrbk 1 Bascs Suppose a sequence of numbers, say a 0,a 1,a,a 3,... s defned by a certan general relatonshp between, say, three consecutve values of the sequence, e.g. a + +3a +1

More information

On the SO 2 Problem in Thermal Power Plants. 2.Two-steps chemical absorption modeling

On the SO 2 Problem in Thermal Power Plants. 2.Two-steps chemical absorption modeling Internatonal Journal of Engneerng Reearch ISSN:39-689)(onlne),347-53(prnt) Volume No4, Iue No, pp : 557-56 Oct 5 On the SO Problem n Thermal Power Plant Two-tep chemcal aborpton modelng hr Boyadjev, P

More information

Physics 120. Exam #1. April 15, 2011

Physics 120. Exam #1. April 15, 2011 Phyc 120 Exam #1 Aprl 15, 2011 Name Multple Choce /16 Problem #1 /28 Problem #2 /28 Problem #3 /28 Total /100 PartI:Multple Choce:Crclethebetanwertoeachqueton.Anyothermark wllnotbegvencredt.eachmultple

More information

A Result on a Cyclic Polynomials

A Result on a Cyclic Polynomials Gen. Math. Note, Vol. 6, No., Feruary 05, pp. 59-65 ISSN 9-78 Copyrght ICSRS Pulcaton, 05.-cr.org Avalale free onlne at http:.geman.n A Reult on a Cyclc Polynomal S.A. Wahd Department of Mathematc & Stattc

More information

Feature Selection: Part 1

Feature Selection: Part 1 CSE 546: Machne Learnng Lecture 5 Feature Selecton: Part 1 Instructor: Sham Kakade 1 Regresson n the hgh dmensonal settng How do we learn when the number of features d s greater than the sample sze n?

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Multiple-objective risk-sensitive control and its small noise limit

Multiple-objective risk-sensitive control and its small noise limit Avalable onlne at www.cencedrect.com Automatca 39 (2003) 533 541 www.elever.com/locate/automatca Bref Paper Multple-objectve rk-entve control and t mall noe lmt Andrew E.B. Lm a, Xun Yu Zhou b; ;1, John

More information

arxiv: v1 [cs.gt] 15 Jan 2019

arxiv: v1 [cs.gt] 15 Jan 2019 Model and algorthm for tme-content rk-aware Markov game Wenje Huang, Pham Vet Ha and Wllam B. Hakell January 16, 2019 arxv:1901.04882v1 [c.gt] 15 Jan 2019 Abtract In th paper, we propoe a model for non-cooperatve

More information

Confidence intervals for the difference and the ratio of Lognormal means with bounded parameters

Confidence intervals for the difference and the ratio of Lognormal means with bounded parameters Songklanakarn J. Sc. Technol. 37 () 3-40 Mar.-Apr. 05 http://www.jt.pu.ac.th Orgnal Artcle Confdence nterval for the dfference and the rato of Lognormal mean wth bounded parameter Sa-aat Nwtpong* Department

More information

CHAPTER 9 LINEAR MOMENTUM, IMPULSE AND COLLISIONS

CHAPTER 9 LINEAR MOMENTUM, IMPULSE AND COLLISIONS CHAPTER 9 LINEAR MOMENTUM, IMPULSE AND COLLISIONS 103 Phy 1 9.1 Lnear Momentum The prncple o energy conervaton can be ued to olve problem that are harder to olve jut ung Newton law. It ued to decrbe moton

More information

bounds compared to SB and SBB bounds as the former two have an index parameter, while the latter two

bounds compared to SB and SBB bounds as the former two have an index parameter, while the latter two 1 Queung Procee n GPS and PGPS wth LRD Traffc Input Xang Yu, Ian L-Jn Thng, Yumng Jang and Chunmng Qao Department of Computer Scence and Engneerng State Unverty of New York at Buffalo Department of Electrcal

More information

Distributed Control for the Parallel DC Linked Modular Shunt Active Power Filters under Distorted Utility Voltage Condition

Distributed Control for the Parallel DC Linked Modular Shunt Active Power Filters under Distorted Utility Voltage Condition Dtrbted Control for the Parallel DC Lnked Modlar Shnt Actve Power Flter nder Dtorted Utlty Voltage Condton Reearch Stdent: Adl Salman Spervor: Dr. Malabka Ba School of Electrcal and Electronc Engneerng

More information

A Two-Stage Modeling and Solution Framework for Multisite Midterm Planning under Demand Uncertainty

A Two-Stage Modeling and Solution Framework for Multisite Midterm Planning under Demand Uncertainty Ind. Eng. Chem. Re. 2000, 39, 3799-3813 3799 A Two-Stage Modelng and Soluton Framework for Multte Mdterm Plannng under Demand Uncertanty Anhuman Gupta and Cota D. Marana* Department of Chemcal Engneerng,

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

Not at Steady State! Yes! Only if reactions occur! Yes! Ideal Gas, change in temperature or pressure. Yes! Class 15. Is the following possible?

Not at Steady State! Yes! Only if reactions occur! Yes! Ideal Gas, change in temperature or pressure. Yes! Class 15. Is the following possible? Chapter 5-6 (where we are gong) Ideal gae and lqud (today) Dente Partal preure Non-deal gae (next tme) Eqn. of tate Reduced preure and temperature Compreblty chart (z) Vapor-lqud ytem (Ch. 6) Vapor preure

More information

Lecture 20: Lift and Project, SDP Duality. Today we will study the Lift and Project method. Then we will prove the SDP duality theorem.

Lecture 20: Lift and Project, SDP Duality. Today we will study the Lift and Project method. Then we will prove the SDP duality theorem. prnceton u. sp 02 cos 598B: algorthms and complexty Lecture 20: Lft and Project, SDP Dualty Lecturer: Sanjeev Arora Scrbe:Yury Makarychev Today we wll study the Lft and Project method. Then we wll prove

More information

No! Yes! Only if reactions occur! Yes! Ideal Gas, change in temperature or pressure. Survey Results. Class 15. Is the following possible?

No! Yes! Only if reactions occur! Yes! Ideal Gas, change in temperature or pressure. Survey Results. Class 15. Is the following possible? Survey Reult Chapter 5-6 (where we are gong) % of Student 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% Hour Spent on ChE 273 1-2 3-4 5-6 7-8 9-10 11+ Hour/Week 2008 2009 2010 2011 2012 2013 2014 2015 2017 F17

More information

Foundations of Arithmetic

Foundations of Arithmetic Foundatons of Arthmetc Notaton We shall denote the sum and product of numbers n the usual notaton as a 2 + a 2 + a 3 + + a = a, a 1 a 2 a 3 a = a The notaton a b means a dvdes b,.e. ac = b where c s an

More information

Order-Based Cost Optimization in Assemble-to-Order Systems

Order-Based Cost Optimization in Assemble-to-Order Systems OPERATIONS RESEARCH Vol. 53, No. 1, January February 2005, pp. 151 169 ssn 0030-364X essn 1526-5463 05 5301 0151 nforms do 10.1287/opre.1040.0146 2005 INFORMS Order-Based Cost Optmzaton n Assemble-to-Order

More information

Hidden Markov Models & The Multivariate Gaussian (10/26/04)

Hidden Markov Models & The Multivariate Gaussian (10/26/04) CS281A/Stat241A: Statstcal Learnng Theory Hdden Markov Models & The Multvarate Gaussan (10/26/04) Lecturer: Mchael I. Jordan Scrbes: Jonathan W. Hu 1 Hdden Markov Models As a bref revew, hdden Markov models

More information

MULTISTART OPTIMIZATION WITH A TRAINABLE DECISION MAKER FOR AVOIDING HIGH-VALUED LOCAL MINIMA

MULTISTART OPTIMIZATION WITH A TRAINABLE DECISION MAKER FOR AVOIDING HIGH-VALUED LOCAL MINIMA 3 rd Internatonal Conference on Experment/Proce/Sytem Modelng/Smulaton & Optmzaton 3 rd IC-EpMO Athen, 8- July, 2009 IC-EpMO MULTISTART OPTIMIZATION WITH A TRAINABLE DECISION MAKER FOR AVOIDING HIGH-VALUED

More information

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming

EEL 6266 Power System Operation and Control. Chapter 3 Economic Dispatch Using Dynamic Programming EEL 6266 Power System Operaton and Control Chapter 3 Economc Dspatch Usng Dynamc Programmng Pecewse Lnear Cost Functons Common practce many utltes prefer to represent ther generator cost functons as sngle-

More information

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

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

More information

A Simple Inventory System

A Simple Inventory System A Smple Inventory System Lawrence M. Leems and Stephen K. Park, Dscrete-Event Smulaton: A Frst Course, Prentce Hall, 2006 Hu Chen Computer Scence Vrgna State Unversty Petersburg, Vrgna February 8, 2017

More information

ORDER FILL RATE, LEADTIME VARIABILITY, AND ADVANCE DEMAND INFORMATION IN AN ASSEMBLE-TO-ORDER SYSTEM

ORDER FILL RATE, LEADTIME VARIABILITY, AND ADVANCE DEMAND INFORMATION IN AN ASSEMBLE-TO-ORDER SYSTEM ORDER FILL RATE, LEADTIME VARIABILITY, AND ADVANCE DEMAND INFORMATION IN AN ASSEMBLE-TO-ORDER SYSTEM YINGDONG LU IBM Research Dvson, T. J. Watson Research Center, Yorktown Heghts, New York 1598, yngdong@us.bm.com

More information

Module 9. Lecture 6. Duality in Assignment Problems

Module 9. Lecture 6. Duality in Assignment Problems Module 9 1 Lecture 6 Dualty n Assgnment Problems In ths lecture we attempt to answer few other mportant questons posed n earler lecture for (AP) and see how some of them can be explaned through the concept

More information

The Essential Dynamics Algorithm: Essential Results

The Essential Dynamics Algorithm: Essential Results @ MIT maachuett nttute of technology artfcal ntellgence laboratory The Eental Dynamc Algorthm: Eental Reult Martn C. Martn AI Memo 003-014 May 003 003 maachuett nttute of technology, cambrdge, ma 0139

More information

Randomness and Computation

Randomness and Computation Randomness and Computaton or, Randomzed Algorthms Mary Cryan School of Informatcs Unversty of Ednburgh RC 208/9) Lecture 0 slde Balls n Bns m balls, n bns, and balls thrown unformly at random nto bns usually

More information

A Novel Approach for Testing Stability of 1-D Recursive Digital Filters Based on Lagrange Multipliers

A Novel Approach for Testing Stability of 1-D Recursive Digital Filters Based on Lagrange Multipliers Amercan Journal of Appled Scence 5 (5: 49-495, 8 ISSN 546-939 8 Scence Publcaton A Novel Approach for Tetng Stablty of -D Recurve Dgtal Flter Baed on Lagrange ultpler KRSanth, NGangatharan and Ponnavakko

More information

Affine transformations and convexity

Affine transformations and convexity Affne transformatons and convexty The purpose of ths document s to prove some basc propertes of affne transformatons nvolvng convex sets. Here are a few onlne references for background nformaton: http://math.ucr.edu/

More information

Optimal Order Policy for Time-Dependent. Deteriorating Items in Response to Temporary. Price Discount Linked to Order Quantity

Optimal Order Policy for Time-Dependent. Deteriorating Items in Response to Temporary. Price Discount Linked to Order Quantity Appled Mathematcal Scence, Vol. 7, 01, no. 58, 869-878 HIKARI Ltd, www.m-hkar.com Optmal Order Polcy for me-dependent Deteroratng Item n Repone to emporary Prce Dcount Lnked to Order uantty R. P. rpath

More information

GREY PREDICTIVE PROCESS CONTROL CHARTS

GREY PREDICTIVE PROCESS CONTROL CHARTS The 4th Internatonal Conference on Qualty Relablty Augut 9-th, 2005 Bejng, Chna GREY PREDICTIVE PROCESS CONTROL CHARTS RENKUAN GUO, TIM DUNNE Department of Stattcal Scence, Unverty of Cape Town, Prvate

More information

2.3 Least-Square regressions

2.3 Least-Square regressions .3 Leat-Square regreon Eample.10 How do chldren grow? The pattern of growth vare from chld to chld, o we can bet undertandng the general pattern b followng the average heght of a number of chldren. Here

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

Optimal Dual-Connectivity Traffic Offloading in Energy-Harvesting Small-Cell Networks

Optimal Dual-Connectivity Traffic Offloading in Energy-Harvesting Small-Cell Networks Th artcle ha been accepted for publcaton n a future ue of th journal, but ha not been fully edted. Content may change pror to fnal publcaton. Ctaton nformaton: DOI.9/TGCN.28.285843, IEEE Tranacton on Green

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

APPLICATIONS OF RELIABILITY ANALYSIS TO POWER ELECTRONICS SYSTEMS

APPLICATIONS OF RELIABILITY ANALYSIS TO POWER ELECTRONICS SYSTEMS APPLICATIONS OF RELIABILITY ANALYSIS TO POWER ELECTRONICS SYSTEMS Chanan Sngh, Fellow IEEE Praad Enjet, Fellow IEEE Department o Electrcal Engneerng Texa A&M Unverty College Staton, Texa USA Joydeep Mtra,

More information

A Robust Method for Calculating the Correlation Coefficient

A Robust Method for Calculating the Correlation Coefficient A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal

More information

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law:

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law: CE304, Sprng 2004 Lecture 4 Introducton to Vapor/Lqud Equlbrum, part 2 Raoult s Law: The smplest model that allows us do VLE calculatons s obtaned when we assume that the vapor phase s an deal gas, and

More information

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017

U.C. Berkeley CS294: Beyond Worst-Case Analysis Luca Trevisan September 5, 2017 U.C. Berkeley CS94: Beyond Worst-Case Analyss Handout 4s Luca Trevsan September 5, 07 Summary of Lecture 4 In whch we ntroduce semdefnte programmng and apply t to Max Cut. Semdefnte Programmng Recall that

More information

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

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

More information

A new construction of 3-separable matrices via an improved decoding of Macula s construction

A new construction of 3-separable matrices via an improved decoding of Macula s construction Dscrete Optmzaton 5 008 700 704 Contents lsts avalable at ScenceDrect Dscrete Optmzaton journal homepage: wwwelsevercom/locate/dsopt A new constructon of 3-separable matrces va an mproved decodng of Macula

More information

Computer Control Systems

Computer Control Systems Computer Control ytem In th chapter we preent the element and the bac concept of computercontrolled ytem. The dcretaton and choce of amplng frequency wll be frt examned, followed by a tudy of dcrete-tme

More information

TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES

TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES TAIL BOUNDS FOR SUMS OF GEOMETRIC AND EXPONENTIAL VARIABLES SVANTE JANSON Abstract. We gve explct bounds for the tal probabltes for sums of ndependent geometrc or exponental varables, possbly wth dfferent

More information

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results.

For now, let us focus on a specific model of neurons. These are simplified from reality but can achieve remarkable results. Neural Networks : Dervaton compled by Alvn Wan from Professor Jtendra Malk s lecture Ths type of computaton s called deep learnng and s the most popular method for many problems, such as computer vson

More information

The optimal delay of the second test is therefore approximately 210 hours earlier than =2.

The optimal delay of the second test is therefore approximately 210 hours earlier than =2. THE IEC 61508 FORMULAS 223 The optmal delay of the second test s therefore approxmately 210 hours earler than =2. 8.4 The IEC 61508 Formulas IEC 61508-6 provdes approxmaton formulas for the PF for smple

More information

Adaptive Memory Programming for the Robust Capacitated International Sourcing Problem

Adaptive Memory Programming for the Robust Capacitated International Sourcing Problem Adaptve Memory Programmng for the Robut Capactated Internatonal Sourcng Problem Joé Lu González Velarde Centro de Stema de Manufactura, ITESM Monterrey, Méxco Lugonzal@campu.mty.tem.mx Rafael Martí Departamento

More information

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations

9 Derivation of Rate Equations from Single-Cell Conductance (Hodgkin-Huxley-like) Equations Physcs 171/271 - Chapter 9R -Davd Klenfeld - Fall 2005 9 Dervaton of Rate Equatons from Sngle-Cell Conductance (Hodgkn-Huxley-lke) Equatons We consder a network of many neurons, each of whch obeys a set

More information