A MULTI-OBJECTIVE GENETIC ALGORITHM FOR EXTEND

Size: px
Start display at page:

Download "A MULTI-OBJECTIVE GENETIC ALGORITHM FOR EXTEND"

Transcription

1 A MULTI-OBJECTIVE GENETIC ALGORITHM FOR EXTEND Bran Kernan and John Geraghty The School of Mechancal & Manufacturng Engneerng, Dubln Cty Unversty, Dubln 9, Ireland. Correspondence: Emal: Abstract: Evolutonary-algorthms are partcularly suted for the task of searchng decson spaces, because they process a set of solutons n parallel, and explot smlartes of pareto solutons through recombnaton. The man part of ths work was to construct a pareto-optmal genetc-algorthm n Extend; a smulaton computer software package that can be used to model manufacturng systems. The algorthm developed, s a robust searchng tool that can be appled to a wde varety of systems, and can analyse decson spaces comprsed of both ntegers and real numbers. A four-stage tandem producton lne as gven n [6] was then modelled n Extend for three dfferent control polces; kanban, CONWIP, and hybrd kanban-conwip. The system had two objectves; maxmse servce level whle mnmsng the work-n-progress nventory. The decson spaces were mapped out, and the pareto-optmser then searched for the pareto-optmal solutons.e. those sets of system parameters, whch gve non-domnated objectve vectors. For the manufacturng system wth the kanban control polcy, t was found that the genetc-algorthm had a maxmum effcency ndex of 3.71 for fndng 40% of the non-domnated solutons, wth a mutaton rate of %. Ths meant that the Genetc Algorthm (GA) s 3.71 tmes more effectve than searchng the decson space randomly. The hybrd control polcy only had a maxmum effcency ndex of 1.88 for fndng 40% of the nondomnated solutons, wth a mutaton rate of %. A possble reason for the reduced effcency when compared wth the kanban control polcy could be due to the reduced confdence levels n the objectves for the decson varables. There s possbly a large potental for mprovng the genetc-algorthm. Nchng and non-nchng technques could be ntroduced, and dfferent selecton, recombnaton and mutaton technques could be used. Comparatve studes should also be carred out wth other mult-objectve genetc algorthms such as the vector evaluated genetc algorthm (VEGA) [3] and the non-domnated sortng algorthm (NSGA) [7]. Keywords: Pareto-Optmal Mult Objectve Genetc Algorthm, Extend, Producton Control Strateges, CONWIP, Kanban,.Kanban-CONWIP Hybrd. 1.0 Introducton Many real-world problems nvolve smultaneous optmsaton of multple objectves that are often competng. One way to solve mult-objectve problems s to transform the orgnal problem nto a sngle-objectve one, by weghtng the objectves wth a weght vector [1]. Ths process allows the use of any sngle-objectve optmsaton algorthm. The obtaned soluton however, depends on the weght vector used n the weghtng process. Choosng the weght vector means that some pror knowledge about the soluton or decson space s requred. Another mult-objectve optmsaton technque that does not mpose an ll-nformed weghtng process on the task of selectng a sngle optmal soluton s pareto-optmsaton. Pareto-optmsaton ams to fnd the set of non-domnated soluton ponts,.e. those vectors for whch ther components cannot be all smultaneously mproved wthout havng a detrmental effect on at least one of the remanng 83

2 components. Ths s known as the concept of pareto-optmalty. Mathematcally mult-objectve problems can be wrtten generally as [1]: Maxmse or Mnmse Subject to: x = { x,x,...,x } X f (x) where: y { f (x),f (x),...,f (x)} Y 1 N = for { 1,...,M} 1 M The vector of decson varables s x of sze N, and X s the decson space. The vector of objectves s y of sze M, and Y s called the objectve space. The soluton to a mult-objectve problem s usually no unque, but a set of equally effcent, non-nferor or non-domnated solutons; the pareto-optmal set. Mathematcally n the maxmsaton case, we can say that the soluton x 1 domnates x or s superor to x (wrtten: x 1 f x ) ff: { 1,...,M}, (x ) y(x ) y { 1,...,M} y (x ) y (x ) 1 1 > If any other x n the decson space X does not domnate x 1, then x 1 s a non-domnated pont. Two nondomnated ponts are ndfferent to each other. A pareto-optmsaton algorthm should search the decson space, fndng the entre set of pareto-optmal solutons. Gven below s a smple multobjectve problem: Maxmse: 1 ( 4 f (x) = 5 x ) Mnmse: f ( x) = 6 + ( x ) 5 Subject to: x R One way to solve ths problem s to create a sngle objectve functon. For example: Maxmse: s(x) = f1(x) 5 = f (x) 6 + ( x 4) ( x 5) Subject to: x R Fgure 1: Sngle-objectve functon Fgure : Pareto-optmal set An optmsaton algorthm can then be appled to ths sngle objectve functon whch wll result n a sngle soluton; x = If a pareto-optmsng algorthm s to be used, then the goal s to fnd the set of solutons such that: { } x, xj X f ( x) f ( xj) f ( x) f ( xj) f ( x) > f ( xj) f ( x) < f ( xj) All the x s that satsfy the above condton belong to the pareto-optmo set, P. For ths example the soluton set P would be all the real numbers n the range [4,5]. Shown n Fgures 1 and above are the graphcal solutons to the mult-objectve problem of maxmsng one objectve, whle tryng to mnmse the other. Fgure 1 shows the soluton, as would be obtaned usng a sngle-objectve optmsaton algorthm, whle Fgure shows the pareto-optmo soluton set. 84

3 In order to obtan the pareto-optmal set of solutons for a mult-objectve optmsaton problem, the decson space X needs to be explored usng a search algorthm. One such search algorthm that can be used s the genetc-algorthm (GA). GAs mmc the metaphor of natural Darwnan bologcal evoluton. They operate on a populaton of potental solutons applyng the prncple of survval of the fttest to produce successvely better approxmatons to a soluton. Evolutonary algorthms dffer substantally from more tradtonal search and optmsaton methods. The most sgnfcant dfferences are []: They search a populaton of ponts n parallel. They do not requre dervatve nformaton or other auxlary knowledge; only the objectve functon and correspondng ftness levels nfluence the drectons of search. They use probablstc transton rules, not determnstc. They are generally more straghtforward to apply, because no restrctons for the defnton of the objectve functon exst. They can provde a number of potental solutons to a gven problem. The fnal choce s left to the user. (Thus, n cases where the partcular problem does not have one ndvdual soluton, for example a famly of pareto-optmal solutons, as n the case of mult-objectve optmsaton and schedulng problems, then the evolutonary algorthm s potentally useful for dentfyng these alternatve solutons smultaneously.) GAs have been recognsed as well suted to mult-objectve optmsaton. Mult-objectve genetcalgorthms (MOGAs) are apt to searchng decson spaces whch are nosy, dscontnuous, mpractcal to search exhaustvely and contan no dervatve nformaton [1]. The GAs used for mult objectve optmsaton problems are generally smlar to a sngle objectve genetc algorthm n everyway, apart from the ftness evaluaton and selecton mechansm. The poneerng work n genetc-algorthms for mult-objectve problems was conducted by Schaffer [3]. The author developed a program named vector evaluated genetc algorthm (VEGA) to fnd paretooptmal solutons to mult-objectve problems. Snce a soluton has several objectve functons, these form a vector. For a problem wth n objectve functons, a populaton s dvded nto n subpopulatons, each of whch has the same number of ndvduals. Every subpopulaton s assocated wth one objectve functon. The solutons n a subpopulaton are evaluated usng only the assocated objectve. Selecton s performed ndependently on each of the subpopulatons, but crossover s carred out across subpopulaton boundares. Whle mplementaton s relatvely straght forward, VEGA has a strong tendency to converge on extreme ponts where only one crteron s optmsed. In other words, the algorthm has lmtatons n fndng solutons n the mddle of extremes. In order to overcome the problems assocated wth VEGA, Goldberg [4] suggested the use of a pareto optmalty rankng method called a pareto GA. The ftness of a set of ndvduals (strng) s based on the rank, whch s determned by usng a non-domnated sortng procedure. Ths procedure fnds the set of strngs that s pareto optmal n the current populaton. The hghest rank s assgned to the strngs, and then the strngs are removed from further contenton. The pareto optmal strngs from the remanng populaton are assgned the next hghest rank. The process contnues untl all the strngs are ranked. After a rank for every strng s determned, a rankng selecton scheme s appled. Another nterestng approach was proposed by Horn et al. [5]. Ths method attempts to dstrbute the populaton evenly along non-domnated fronters. Snce ths approach s based on the concept of nchng, t s named nched pareto GA. The prmal selecton s carred out usng bnary tournament adapted to control domnaton pressure. Two canddates and a comparson set are pcked at random from the populaton. If one canddate s domnated by the comparson set and the other s not, the latter s selected. If nether are or both are non-domnated by the comparson set, then a ftness sharng mechansm comes nto play. The ftness of an ndvdual s degraded by the densty of the ndvduals wthn a certan dstance from t. The sphercal regon formed by the dstance s called the nche. Snce the amount of degradaton s nversely proportonal to the number of ndvduals n a nche, a sparser nche ncreases the resultng ftness values. As a consequence, convergence nto a narrow regon of soluton space s prevented. 85

4 .0 The Pareto-Optmal Genetc Algorthm The pareto-optmal genetc algorthm developed for Extend ams to fnd the set of non-domnated solutons n a decson space comprsed of up to 8 decson varables, for an objectve vector comprsed of up to 6 components. The decson space can consst of numbers n nteger or real number format encoded as decmals. The algorthm works as follows: 1. Intalsaton An ntal populaton of 5 ndvduals s created at random from the decson space.. Domnance Check The current populaton s checked to see whch ndvduals are non-domnated. Mathematcally n the maxmsaton case, we say that ndvdual x 1 domnates x or s superor to x (wrtten: x 1 f x ) ff: { 1,...,M}, (x ) y(x ) y { 1,...,M} y (x ) y (x ) 1 1 > If any other x n the decson space X does not domnate x 1, then x 1 s a non-domnated pont. Two nondomnated ponts are ndfferent to each other. Indvduals whch are non-domnated are assgned a boolean value of true. 3. Adjustment Any ndvduals n the current populaton, whch are domnated, are replaced wth an offsprng created from two dfferent non-domnated parents. The parents are obtaned at random from the current populaton, wth equal probablty of beng chosen. If there s only one non-domnated soluton however (there wll also be at least one), then the second parent s obtaned at random from the decson space. Because the decson space can be nteger or real format, t was decded to use a hybrd recombnaton technque of dscrete and ntermedate recombnaton, whch would be referred to as trad-based recombnaton (TBR). Wth TBR, each chromosome that makes up the offsprng, can be obtaned three dfferent ways wth equal probablty: t can take on the value of the frst parent s chromosome, the value of the second parent s chromosome, or t can take on a value n between, and ncludng the two parents chromosome values. When each chromosome s created, exo-parental mutaton [] can occur, wth a probablty of occurrence defned by the user. 4. Domnance Recheck The adjusted populaton s rechecked, to see whch solutons are non-domnated. 5. Increase Populaton If all of the ndvduals of the current populaton are non-domnated, 5 new offsprng are created and added to the populaton. The offsprng are created n the same manner as descrbed n step 3 of the genetc-algorthm. 6. Repeat The algorthm s repeated by gong back to step. The algorthm can be termnated by the number of generatons or the number of deadruns specfed by the user. The algorthm wll also end f the entre decson space has been searched (vald f decson space contans nteger values only). 3.0 Verfcaton of Algorthm Once the pareto-optmser had been bult, debugged and verfed on two smple optmsaton problems, t was then appled to a case-study problem from the lterature on determnng the optmal producton control strategy for a lean manufacturng system. The manufacturng system analysed wth the GA was a four-stage tandem producton lne wth each stage consstng of a machne and an output buffer as descrbed n [6]. The machne operaton tmes were modelled usng a lognormal dstrbuton wth mean 0.98 mnutes and a standard devaton of 0.0 mnutes. The machnes faled accordng to an exponental dstrbuton wth a mean tme-between-falures of 1000 mnutes. The tme-between-falures measure spent workng tme, not clock tme. Ths s known as operaton dependent falure and mples that no machne can fal unless t s workng on a part. The repar tme of the machnes was also modelled usng an exponental dstrbuton wth a mean of 3 mnutes. The demand rate was constant at 1 part per mnute. Other assumptons across all schedulng polces are: 86

5 Materal s transported n unts of one wthout delay. Informaton such as kanban s transmtted nstantaneously. Machnes operate asynchronously, so parts can be loaded whenever a part s present and the proper authorsaton has been receved. Jobs authorsed for loadng follow a frst come frst served dspatch polcy. Any demand that cannot be satsfed from the fnshed goods nventory s lost to the system. Demands occur constantly, whether or not they are beng satsfed. Ths s somewhat dfferent to the procedure used n [6], where the demand process comes to a stop untl each demand s satsfed. The objectves to be optmsed were the servce level (maxmsed) and work-n-progress (WIP) (mnmsed). Servce level s the rato of met demands to total demands and deally should be 1.0. The WIP s the total amount of nventory n the system. To reduce costs, the WIP should be kept as low as possble. These two parameters have a curve-lnear relatonshp, wth the servce level ncreasng as the amount of nventory (WIP) n the system ncreases. 3.1 Optmsaton and Decson Space for Kanban For the manufacturng system wth a kanban control polcy, cells 1, and 3 could take on kanban values n the range 1 to 5, whle cell 4 could take on kanban values n the range 1 to 10. Ths gave a total of 1,50 cases,.e. 1,50 dscrete ponts n the decson space X. The mult-objectve genetcalgorthm therefore, ams to obtan the set of pareto-optmal solutons (cell kanban) so that: { S(N ) S(N ) W(N ) W(N ) [ S(N ) > S(N ) W(N ) W(N )]} N,Nj X j j j < N = { n1,n,n3, n4 }... Cell Kanban S (N).. Servce Level for N W (N) Work-In-Progress for N j Fgure 3: Soluton space for kanban manufacturng system The entre decson space was mapped out by modelng the system usng Extend, wth 0 replcatons for 30,000 mnutes, and a warm-up perod of 5,000 mnutes, for each of the 1,50 cases. Servce levels were only calculated after the warm-up perod had fnshed, and the WIP was averaged between 5,000 and 30,000 mnutes. The decson space was placed n a data fle, so the objectve vectors for the dfferent soluton sets could be easly obtaned. The other alternatve was havng to run the smulaton to obtan servce levels and WIP levels as the genetc-algorthm demanded. By havng the decson space mapped out, the analyss and verfcaton of the genetc-algorthm on the manufacturng system was sgnfcantly qucker. Accordng to [6], the system should deally be run wth replcatons for 10,000 mnutes, wth a warm-up perod of 9,600 mnutes, to gve accurate results. Accurate results n ths case meanng, 95% confdence nterval half-wdths of for the servce level and 0.04 for the WIP would be requred. Due to computer lmtatons however, confdence level half-wdths of ths sze would have taken too long to obtan. The am of these experments however was to verfy the GA, 87

6 rather than accurately solve a manufacturng problem. The soluton space s shown n Fgure 3 for servce levels of greater than 99.9%. A hermte splne was ftted to the non-domnated solutons to show the pareto-front. In total t was found that there were 59 pareto solutons n the entre decson space. The decson space was then placed n a MATLAB data fle, and the algorthm was translated to MATLAB so a more extensve analyss of the GA could be carred out. 3. Evaluaton of the Performance of the Algorthm for Kanban In order to examne the effectveness of the genetc-algorthm on the manufacturng system, an ndex of effcency was defned. The effcency ndex E s smply the rato of the percentage of decson space searched, to the percentage of the pareto set obtaned. For example, f 50% of the decson space s searched randomly, then approxmately 50% of the pareto set wll be obtaned. Ths would gve an effcency ndex of 1.0. Before the genetc-alogrthm was appled to the problem, a stochastc paretooptmser (SPO) was frst used. The SPO works n a smlar fashon to the genetc-algorthm, except that new chldren are obtaned randomly from the decson space. The SPO was run for 150 generatons (obtans approxmately 3% pareto-set) wth 50 replcatons. The effcency ndex was found to be ± wth 90% confdence ntervals. Ths value of effcency ndex s consstent wth theory. Table 1: Non-domnated soluton set for manufacturng system wth kanban control Kanban Kanban Kanban Kanban Kanban Kanban Kanban Kanban Fgure 4: Effcency Vs. Mutaton Fgure 5: Effcency Vs. Pareto % The genetc-algorthm should have an effcency ndex greater than one, because t uses smple artfcal ntellgence to search the space. If there s a pattern to the pareto-optmal soluton set, then the algorthm wll fnd t. In ths case, t would be expected that the non-domnated solutons would have a hgher number of kanban n the fourth cell, then n the other three cells (see Table 1). Fgure 4 shows the effcency ndces for fndng 70% of the pareto-set wth dfferent mutaton rates. Fgure 5 shows the effcency ndces of the genetc-algorthm for fndng dfferent percentages of the pareto-set, wth a % mutaton rate. A % mutaton rate was chosen because accordng to the data n Fgure 4 t gves the largest effcency ndex for fndng 70% of the pareto-set. The effcency ndces were calculated by runnng the genetc-algorthm for 100 generatons. Confdence ntervals (90%) for the effcency ndces are ndcated by the arrows on the fgures. In both graphs, the two sets of data have been ftted wth a four-pont hermte splne, to show the general trend of the effcency ndex as a functon of mutaton rate and pareto-set percentage. 3.3 Dscusson of Performance of the Algorthm on Kanban System Accordng to the data n Fgure 4, the effcency of the algorthm peaks at, approxmately, 3.38 wth a mutaton rate of %, and then steadly decreases. Ths s consstent wth theory; a mutaton rate that s 88

7 too low, means that the genetc-algorthm can get stuck n a local set of solutons, and a mutaton rate that s too hgh can adversely affect the breedng of new solutons. For most pareto-optmal genetcalgorthms, mutaton rates are generally between 1% and 30%. It should be noted however, that the best mutaton rate s dependent on the actual problem. Also, for ths problem a mutaton rate of % gves the hghest effcency for fndng 70% of the pareto set. It may be the case, that dfferent mutaton rates gve hgher effcences for fndng dfferent percentages of the pareto set. Accordng to the data n Fgure 5, the effcency steadly ncreases from 10% of the pareto-set and reaches a peak at 40% of the pareto-set, and then steadly decreases. Agan ths s consstent wth theory. The effcency s low for fndng 10% of the pareto-set, because the genetc-algorthm does not have enough ft ndvduals yet to determne the complete pattern to the pareto-set (f there s any). The effcency peaks at 40%, ndcatng that 40% of the pareto-set follows a pattern that the genetcalgorthm can fnd. The remnng 60% of the set starts to follow a more random pattern, and so the effcency of the genetc-algorthm decreases. The effcency for fndng the entre non-domnated soluton set s 1.6,.e. 6% more effcent then searchng the decson space randomly. The reason ths value s so low, s because the pareto soluton set s obtaned from a system wth stochastc nputs, and therefore some solutons wll probably be completely anomalous and random; consequently t s very dffcult for the genetc-algorthm to fnd these solutons. It s mportant to note that the system was only run for 30,000 mnutes wth 0 replcatons. Ths may not gve suffcent confdence for the pareto-soluton set. Orgnally the system was run for 30,000 mnutes wth 10 replcatons and gave only 47 non-domnated solutons (compared wth 59 non-domnated solutons for 30,000 mnutes wth 0 replcatons, see Table 1). In theory, runnng the system for a longer tme and wth addtonal replcatons, wll gve more confdence for the pareto soluton set, and consequently the effcency of the genetc-algorthm should ncrease. Another way to check the valdty of the genetc-algorthm s to run the system wth dfferent control polces such as CONWIP, and hybrd kanban-conwip, as mentoned n [6], and to use the MOGA to obtan the pareto-soluton set for these control polces. 3.4 Applcaton of the Algorthm for Hybrd Polcy A smlar analyss was conducted for the kanban-conwip hybrd producton control strategy. Ths producton control strategy dstrbutes Kanbans to ndvdual workcentres and places a lmt, known as a WIP Cap, on the total amount of nventory that may be n the system at any gven pont n tme. Once ths lmt has been reached nventory may not enter the system untl a demand event removes a corespondng amount of nventory from the system. Overall, the genetc-algorthm was neffcent at searchng the hybrd decson space. The hghest effcency obtaned was only 1.88 for obtanng 40% of the pareto set. In order to obtan 100% of the pareto set, the effcency was only 1.1,.e. 89% of the decson space needed to be searched. The effcency ndces are so low because the genetc-algorthm was unable to fnd a pattern among the pareto optmal solutons. The reason that there appears to be no pattern to the pareto soluton set, s probably because there was nsuffcent confdence n the values for servce level and WIP n the decson space. An mportant observaton, s that some of the ndvduals n the non-domanted soluton set, do not make sense from a practcal vewpont. For nstance the frst non-domnated soluton found had a cap of 1, and 3 kanbans crculatng n the frst cell; there s no need to have 3 kanband crculatng n a cell, snce only one part s allowed to be n the system at any one tme. In fact, there were a total of 4 ndvduals out of the 97 non-domnated ndvduals.e. 5%, that have kanban values n ndvdual cells that are hgher than the cap for the system. If these ndvduals are removed from the pareto soluton set, and f the genetc-algorthm s told to gnore ndvduals whch have cell kanban hgher than the cap, then the effcency ndex for fndng 40% of the pareto-set s now.1 ± Although ths s only a margnal mprovement, t s an mportant consderaton when analysng manufacturng systems wth hybrd kanban-conwip control polces; the cap should always be greater than or equal to the ndvdual kanban values for each cell. In theory, runnng the system for more replcatons and/or a longer smulaton tme, wll gve more confdence for the pareto soluton set, and consequently the effcency of the genetc-algorthm should ncrease. In order to reduce processng tme, the ndvduals n the decson space that have cell kanbans hgher than the cap can be gnored. Ths means that the decson space of 3,15 ndvduals s reduced to,75 ndvduals (a 13% decrease). However, t may be the case that the ndvduals that make up the pareto soluton set wll not have a dscernable pattern that the genetc-algorthm can easly fnd. 89

8 Regardless, the genetc-algorthm stll operates up to twce as effectvely as a SPO would, for ths system wth a hybrd control polcy. 3.5 Comparson of Control Strateges Shown below n Fgure 6 are the pareto-fronts for the 4-stage tandem producton wth the three dfferent control polces; kanban, CONWIP and hybrd. The fronts were obtaned by fttng hermte splnes to the non-domnated solutons. The hybrd control polcy appears to be the best schedulng polcy (best n ths case, apples to both servce level and WIP). If a servce level of 99.9% s desred, then the kanban control polcy would result n a WIP of approxmately 13, CONWIP 1 and hybrd only 10. (These values are sgnfcantly dfferent to those obtaned n [6] whch are kanban 15.8, CONWIP, 14.6 and hybrd 13.93). The dfference s probably due to the dfferent demand processes). Essentally the hybrd control polcy s superor to kanban and CONWIP, because t combnes the ndvdual advantages of both: It places a lmt on the total amount of nventory n the system (CONWIP), and uses kanban to ensure that ndvdual buffer levels do not get too hgh when bottleneckng occurs. Kanban CONWIP Hybrd Fgure 6: Pareto-fronts for manufacturng system wth dfferent control polces It s mportant to note that for the kanban and hybrd control polces, the hghest effcency ndces are for fndng 40% of the pareto soluton set (kanban 3.71, hybrd 1.88). If a manufacturng system s beng analysed to see whch s the optmal control polcy, then 40% of the pareto-soluton set, should be suffcent to determne the pareto front. Tryng to fnd the complete pareto soluton set for stochastc systems, wll probably result n reduced effectveness of the genetc-algorthm, due to anomalous ndvduals n the pareto-set. 4.0 Dscusson and Future Work The genetc-algorthm performs reasonably well on the gven problem of optmsng a manufacturng system. Unfortunately, no research papers on pareto-optmal genetc-algorthms appled to ths knd of problem could be found, so no comparatve studes could be carred out. Consequently there s no way to comparatvely quantfy the effectveness of the genetc-algorthm. Further work would nvolve buldng other evolutonary-algorthms such as the vector evaluated genetc algorthm (VEGA) presented by Schaffer [3], and the non-domnated sortng algorthm (NSGA) developed by Srnvas and Deb [7], and dong a comparatve study on the manufacturng system. VEGA conducts selecton for each objectve separately. In detal, the matng pool s dvded nto parts of equal sze and ndvduals are chosen at random from the current populaton accordng to each objectve. Afterwards the matng pool s shuffled and crossover and mutaton are performed as usual. Schaffer mplemented ths method n combnaton wth ftness proportonate selecton. Although some serous drawbacks are known, ths algorthm has been a strong pont of reference n the feld of evolutonary-algorthms. 90

9 Wth NSGA, the ftness assgnment s carred out n several steps. In each, the non-domnated solutons consttutng a non-domnated front are assgned the same dummy ftness value. These solutons are shared wth ther dummy ftness values and gnored n the further classfcaton process. Fnally, the dummy ftness s set to a value less than the smallest shared ftness value n the current non-domnated front. Then the next front s extracted. Ths procedure s repeated untl all ndvduals n the populaton are classfed. In order to get more accurate results from the evolutonary algorthms, mproved confdence n the decson space for the manufacturng system wth kanban and hybrd control polces would also need to be obtaned. By comparng and analysng the dfferent MOGAs t wll probably be possble to buld a more effectve and robust genetc-algorthm. It s mportant to remember that evolutonary algorthms use a wde and vared set of technques and methods and that there s potentally a large scope for mprovng the operaton of the genetc-algorthm. Some of the possble changes to the genetc algorthm nclude: Encodng chromosomes n bnary format Currently, the chromosomes that make up the decson space are encoded n decmal format. The chromosomes could be encoded as bnary numbers usng Gray code or Bnary Coded Decmals (BCDs). Each bnary bt would correspond to the actvaton or deactvaton of a gene. Nchng and non-nchng technques The am of both of these technques s to preserve dversty n the populaton and therefore try to prevent premature convergence. Nchng technques are charactersed by ther capablty to promote the formulaton and mantenance of stable subpopulatons (nches). Ftness sharng s used most frequently, whch s a nchng technque based on the dea that ndvduals n a partcular nche have to share the avalable resources. The more ndvduals are located n the neghborhood of a certan ndvdual, the more ts ftness value s degraded. The neghborhood s defned n terms of a dstance measure and specfed by the so-called nche radus σ share. Currently, most MOGAs mplement ftness sharng. Among the non-nchng technques, restrcted matng s the most common n mult-crtera functon optmsaton. Bascally two ndvduals are allowed to mate only f they are wthn a certan dstance (gven by the parameter σ mate ) to each other. Ths mechansm may avod the formaton of unft ndvduals and therefore mprove the onlne performance. Nevertheless, as mentoned n [8], t does not appear to be wdespread n the feld of mult-objectve evolutonary algorthms. Dfferent selecton processes The selecton process currently used n the genetc-algorthm s farly smplstc, and there s potentally large room for mprovement. One selecton process that could be mplemented s that used n the nched pareto genetc algorthm (NPGA) [5]. The NPGA combnes tournament selecton wth the concept of pareto domnance. Two competng ndvduals and a comparson set of other ndvduals are pcked at random from the populaton; the sze of the comparson set s determned by a comparson parameter. If one of the competng ndvduals s domnated by any member of the set, and the other s not, then the latter s chosen as the wnner of the tournament. If both ndvduals are domnated (or nondomnated), the result of the tournament s decded by sharng: The ndvdual that has the least ndvduals n ts nche s selected for reproducton Dfferent recombnaton processes Dfferent methods of creatng offsprng from parents such as lne recombnaton and mult-pont crossover could be used. The latter recombnaton technque can only be used f the decson space s encoded n bnary format. Dfferent mutaton processes There are a number of dfferent ways the mutaton process could be altered. For example, f the decson space s encoded n bnary format, then mutaton could be allowed to occur on one chromosome per new offsprng created. Also t may be possble to ntroduce a varable mutaton rate that changes throughout an evolutonary run. When dealng wth genetc-algorthms t s mportant that the algorthm not be calbrated to the problem used for valdatng the algorthm. It may be the case that the GA wll work excellently on the 91

10 valdatng problem, but not on any other optmsaton problems. To overcome ths, a wde and vared array of problems should be used to verfy the algorthm. In ths project, the GA needs to be able to search decson spaces wth real numbers as well as ntegers, so t should be tested on large nosy, dscontnuous decson spaces comprsed of real numbers. Genetc-algorthms are currently a fertle ground for research and applcaton development. There are a plethora of dfferent ways a genetcalgorthm can be bult; some are robust and desgned to optmse a wde varety of problem, whle others are desgned for specfc problems. Whle a rch set of technques and models are avalable coverng a range of domans, there are stll many areas n the feld of evolutonary algorthms remanng to be understood and exploted. 5. Conclusons One of the man conclusons to be drawn from ths work s that pareto-optmal MOGAs can be an effcent method of searchng nosy dscontnuous decson spaces, although ther effectveness and robustness s dependent on the ablty of the algorthm to fnd smlartes between non-domnated solutons. The MOGA bult n Extend was tested on a four-stage producton lne for three dfferent control polces; kanban, CONWIP, and hybrd kanban-conwip. For the manufacturng system wth kanban control polcy, t was found that the genetc-algorthm had a maxmum effcency ndex of 3.71 for fndng 40% of the non-domnated solutons, wth a mutaton rate of %. The CONWIP control polcy only had a decson space of 5 varables and 4 of those were non-domnated; consequently t was determned that t s neffcent for a GA to search such a small decson space. The system wth the hybrd control polcy only had a maxmum effcency ndex of 1.88 for fndng 40% of the non-domnated solutons, wth a mutaton rate of %. Because no comparatve studes could be carred out, there s currently way to comparatvely quantfy the effectveness of the genetc-algorthm. However t s expected that the effcency ndces of the MOGA wll ncrease f there s ncreased confdence n the objectve vectors for the decson space. There s large potental for mprovng the genetc-algorthm. Nchng and non-nchng technques could be ntroduced, and dfferent selecton, recombnaton and mutaton technques could be used. Comparatve analyses should also be carred out wth other mult-objectve genetc algorthms such as the vector evaluated genetc algorthm (VEGA) and the non-domnated sortng algorthm (NSGA). References [1] N. Srnvas and K. Deb. Multobjectve Optmzaton usng Nondomnated Sortng n Genetc Algorthms. Tech. Rep., Dept. Mechancal Engneerng, Kanput, Inda, [] Evolutonary Algorthms (accessed 1 st June 04) [3] J. D. Schaffer. Multple Objectve Optmzaton wth Vector Evaluated Genetc Algorthms. Ph.D. dssertaton, Vanderblt Unv., Nashvlle, TN, [4] D. E. Goldberg. Genetc Algorthms n Search, Optmzaton and Machne Learnng. Addson-Wesley, Readng MA, [5] J. Horn, N. Nafplots and D. E. Goldberg. A Nched Pareto Genetc-Algorthm for Mult- Objectve Optmzaton. Proc. of 1 st IEEE-ICEC Conference. 1994, pp [6] A. M. Bonvk, C. E. Couch, and S. B. Gershwn. A Comparson of Producton-Lne Control Mechansms. Internatonal Journal of Producton Research, 35(3):pp , [7] N. Srnvas and K. Deb. Mult-Objectve Optmzaton usng Non-Domnated Sortng n Genetc Algorthms. Evoluton Computng, vol., no. 3, pp. 1 48, [8] C. M. Fonseca and P. J. Flemng. An Overvew of Evolutonary Algorthms n Mult- Objectve Optmzaton. Evoluton Computng, vol. 3, no.1, pp. 1 16,

Smoothing Spline ANOVA for variable screening

Smoothing Spline ANOVA for variable screening Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory

More information

Meta-heuristics for Multidimensional Knapsack Problems

Meta-heuristics for Multidimensional Knapsack Problems 2012 4th Internatonal Conference on Computer Research and Development IPCSIT vol.39 (2012) (2012) IACSIT Press, Sngapore Meta-heurstcs for Multdmensonal Knapsack Problems Zhbao Man + Computer Scence Department,

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

PARETO BAYESIAN OPTIMIZATION ALGORITHM FOR THE MULTIOBJECTIVE 0/1 KNAPSACK PROBLEM

PARETO BAYESIAN OPTIMIZATION ALGORITHM FOR THE MULTIOBJECTIVE 0/1 KNAPSACK PROBLEM PARETO BAYESIAN OPTIMIZATION ALGORITHM FOR THE MULTIOBJECTIVE 0/ KNAPSACK PROBLEM Josef Schwarz Jří Očenáše Brno Unversty of Technology Faculty of Engneerng and Computer Scence Department of Computer Scence

More information

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 A mathematcal programmng approach to the analyss, desgn and

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION 24 CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION The present chapter proposes an IPSO approach for multprocessor task schedulng problem wth two classfcatons, namely, statc ndependent tasks and

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Multiobjective Optimization

Multiobjective Optimization Chapter 10 Multobjectve Optmzaton All men seek one goal: success or happness. The only way to acheve true success s to express yourself completely n servce to socety. Frst, have a defnte, clear, practcal

More information

NGPM -- A NSGA-II Program in Matlab

NGPM -- A NSGA-II Program in Matlab Verson 1.4 LIN Song Aerospace Structural Dynamcs Research Laboratory College of Astronautcs, Northwestern Polytechncal Unversty, Chna Emal: lsssswc@163.com 2011-07-26 Contents Contents... 1. Introducton...

More information

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed

More information

The Codesign Challenge

The Codesign Challenge ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Multi-objective Optimization Using Self-adaptive Differential Evolution Algorithm

Multi-objective Optimization Using Self-adaptive Differential Evolution Algorithm Mult-objectve Optmzaton Usng Self-adaptve Dfferental Evoluton Algorthm V. L. Huang, S. Z. Zhao, R. Mallpedd and P. N. Suganthan Abstract - In ths paper, we propose a Multobjectve Self-adaptve Dfferental

More information

Cracking of the Merkle Hellman Cryptosystem Using Genetic Algorithm

Cracking of the Merkle Hellman Cryptosystem Using Genetic Algorithm Crackng of the Merkle Hellman Cryptosystem Usng Genetc Algorthm Zurab Kochladze 1 * & Lal Besela 2 1 Ivane Javakhshvl Tbls State Unversty, 1, I.Chavchavadze av 1, 0128, Tbls, Georga 2 Sokhum State Unversty,

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

More information

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009. Farrukh Jabeen Algorthms 51 Assgnment #2 Due Date: June 15, 29. Assgnment # 2 Chapter 3 Dscrete Fourer Transforms Implement the FFT for the DFT. Descrbed n sectons 3.1 and 3.2. Delverables: 1. Concse descrpton

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Solving two-person zero-sum game by Matlab

Solving two-person zero-sum game by Matlab Appled Mechancs and Materals Onlne: 2011-02-02 ISSN: 1662-7482, Vols. 50-51, pp 262-265 do:10.4028/www.scentfc.net/amm.50-51.262 2011 Trans Tech Publcatons, Swtzerland Solvng two-person zero-sum game by

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

CHAPTER 4 OPTIMIZATION TECHNIQUES

CHAPTER 4 OPTIMIZATION TECHNIQUES 48 CHAPTER 4 OPTIMIZATION TECHNIQUES 4.1 INTRODUCTION Unfortunately no sngle optmzaton algorthm exsts that can be appled effcently to all types of problems. The method chosen for any partcular case wll

More information

Study on Multi-objective Flexible Job-shop Scheduling Problem considering Energy Consumption

Study on Multi-objective Flexible Job-shop Scheduling Problem considering Energy Consumption Journal of Industral Engneerng and Management JIEM, 2014 7(3): 589-604 nlne ISSN: 2014-0953 Prnt ISSN: 2014-8423 http://dx.do.org/10.3926/jem.1075 Study on Mult-objectve Flexble Job-shop Schedulng Problem

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

A Novel Approach for an Early Test Case Generation using Genetic Algorithm and Dominance Concept based on Use cases

A Novel Approach for an Early Test Case Generation using Genetic Algorithm and Dominance Concept based on Use cases Alekhya Varkut et al, / (IJCSIT) Internatonal Journal of Computer Scence and Informaton Technologes, Vol. 3 (3), 2012,4218-4224 A Novel Approach for an Early Test Case Generaton usng Genetc Algorthm and

More information

Sequential search. Building Java Programs Chapter 13. Sequential search. Sequential search

Sequential search. Building Java Programs Chapter 13. Sequential search. Sequential search Sequental search Buldng Java Programs Chapter 13 Searchng and Sortng sequental search: Locates a target value n an array/lst by examnng each element from start to fnsh. How many elements wll t need to

More information

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

More information

y and the total sum of

y and the total sum of Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton

More information

An Efficient Genetic Algorithm with Fuzzy c-means Clustering for Traveling Salesman Problem

An Efficient Genetic Algorithm with Fuzzy c-means Clustering for Traveling Salesman Problem An Effcent Genetc Algorthm wth Fuzzy c-means Clusterng for Travelng Salesman Problem Jong-Won Yoon and Sung-Bae Cho Dept. of Computer Scence Yonse Unversty Seoul, Korea jwyoon@sclab.yonse.ac.r, sbcho@cs.yonse.ac.r

More information

Optimizing Document Scoring for Query Retrieval

Optimizing Document Scoring for Query Retrieval Optmzng Document Scorng for Query Retreval Brent Ellwen baellwe@cs.stanford.edu Abstract The goal of ths project was to automate the process of tunng a document query engne. Specfcally, I used machne learnng

More information

Problem Set 3 Solutions

Problem Set 3 Solutions Introducton to Algorthms October 4, 2002 Massachusetts Insttute of Technology 6046J/18410J Professors Erk Demane and Shaf Goldwasser Handout 14 Problem Set 3 Solutons (Exercses were not to be turned n,

More information

Programming in Fortran 90 : 2017/2018

Programming in Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values

More information

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

More information

Multi-objective Optimization Using Adaptive Explicit Non-Dominated Region Sampling

Multi-objective Optimization Using Adaptive Explicit Non-Dominated Region Sampling 11 th World Congress on Structural and Multdscplnary Optmsaton 07 th -12 th, June 2015, Sydney Australa Mult-objectve Optmzaton Usng Adaptve Explct Non-Domnated Regon Samplng Anrban Basudhar Lvermore Software

More information

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

More information

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

GA-Based Learning Algorithms to Identify Fuzzy Rules for Fuzzy Neural Networks

GA-Based Learning Algorithms to Identify Fuzzy Rules for Fuzzy Neural Networks Seventh Internatonal Conference on Intellgent Systems Desgn and Applcatons GA-Based Learnng Algorthms to Identfy Fuzzy Rules for Fuzzy Neural Networks K Almejall, K Dahal, Member IEEE, and A Hossan, Member

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

More information

Sorting Review. Sorting. Comparison Sorting. CSE 680 Prof. Roger Crawfis. Assumptions

Sorting Review. Sorting. Comparison Sorting. CSE 680 Prof. Roger Crawfis. Assumptions Sortng Revew Introducton to Algorthms Qucksort CSE 680 Prof. Roger Crawfs Inserton Sort T(n) = Θ(n 2 ) In-place Merge Sort T(n) = Θ(n lg(n)) Not n-place Selecton Sort (from homework) T(n) = Θ(n 2 ) In-place

More information

Simulation Optimization of Pull Control Policies for Serial Manufacturing Lines and Assembly Manufacturing Systems with Genetic Algorithms

Simulation Optimization of Pull Control Policies for Serial Manufacturing Lines and Assembly Manufacturing Systems with Genetic Algorithms Smulaton Optmzaton of Pull Control Polces for Seral Manufacturng Lnes and Assembly Manufacturng Systems wth Genetc Algorthms Dmtros E Koulourots, Alexander S Xanthopoulos, Vaslos D Tourasss To cte ths

More information

NAG Fortran Library Chapter Introduction. G10 Smoothing in Statistics

NAG Fortran Library Chapter Introduction. G10 Smoothing in Statistics Introducton G10 NAG Fortran Lbrary Chapter Introducton G10 Smoothng n Statstcs Contents 1 Scope of the Chapter... 2 2 Background to the Problems... 2 2.1 Smoothng Methods... 2 2.2 Smoothng Splnes and Regresson

More information

Load-Balanced Anycast Routing

Load-Balanced Anycast Routing Load-Balanced Anycast Routng Chng-Yu Ln, Jung-Hua Lo, and Sy-Yen Kuo Department of Electrcal Engneerng atonal Tawan Unversty, Tape, Tawan sykuo@cc.ee.ntu.edu.tw Abstract For fault-tolerance and load-balance

More information

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between

More information

Biostatistics 615/815

Biostatistics 615/815 The E-M Algorthm Bostatstcs 615/815 Lecture 17 Last Lecture: The Smplex Method General method for optmzaton Makes few assumptons about functon Crawls towards mnmum Some recommendatons Multple startng ponts

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory Background EECS. Operatng System Fundamentals No. Vrtual Memory Prof. Hu Jang Department of Electrcal Engneerng and Computer Scence, York Unversty Memory-management methods normally requres the entre process

More information

Classifier Swarms for Human Detection in Infrared Imagery

Classifier Swarms for Human Detection in Infrared Imagery Classfer Swarms for Human Detecton n Infrared Imagery Yur Owechko, Swarup Medasan, and Narayan Srnvasa HRL Laboratores, LLC 3011 Malbu Canyon Road, Malbu, CA 90265 {owechko, smedasan, nsrnvasa}@hrl.com

More information

A Statistical Model Selection Strategy Applied to Neural Networks

A Statistical Model Selection Strategy Applied to Neural Networks A Statstcal Model Selecton Strategy Appled to Neural Networks Joaquín Pzarro Elsa Guerrero Pedro L. Galndo joaqun.pzarro@uca.es elsa.guerrero@uca.es pedro.galndo@uca.es Dpto Lenguajes y Sstemas Informátcos

More information

Optimization of machining fixture layout for tolerance requirements under the influence of locating errors

Optimization of machining fixture layout for tolerance requirements under the influence of locating errors MultCraft Internatonal Journal of Engneerng, Scence and Technology Vol. 2, No. 1, 2010, pp. 152-162 INTERNATIONAL JOURNAL OF ENGINEERING, SCIENCE AND TECHNOLOGY www.jest-ng.com 2010 MultCraft Lmted. All

More information

Private Information Retrieval (PIR)

Private Information Retrieval (PIR) 2 Levente Buttyán Problem formulaton Alce wants to obtan nformaton from a database, but she does not want the database to learn whch nformaton she wanted e.g., Alce s an nvestor queryng a stock-market

More information

A New Token Allocation Algorithm for TCP Traffic in Diffserv Network

A New Token Allocation Algorithm for TCP Traffic in Diffserv Network A New Token Allocaton Algorthm for TCP Traffc n Dffserv Network A New Token Allocaton Algorthm for TCP Traffc n Dffserv Network S. Sudha and N. Ammasagounden Natonal Insttute of Technology, Truchrappall,

More information

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms Course Introducton Course Topcs Exams, abs, Proects A quc loo at a few algorthms 1 Advanced Data Structures and Algorthms Descrpton: We are gong to dscuss algorthm complexty analyss, algorthm desgn technques

More information

Array transposition in CUDA shared memory

Array transposition in CUDA shared memory Array transposton n CUDA shared memory Mke Gles February 19, 2014 Abstract Ths short note s nspred by some code wrtten by Jeremy Appleyard for the transposton of data through shared memory. I had some

More information

A GENETIC ALGORITHM FOR PROCESS SCHEDULING IN DISTRIBUTED OPERATING SYSTEMS CONSIDERING LOAD BALANCING

A GENETIC ALGORITHM FOR PROCESS SCHEDULING IN DISTRIBUTED OPERATING SYSTEMS CONSIDERING LOAD BALANCING A GENETIC ALGORITHM FOR PROCESS SCHEDULING IN DISTRIBUTED OPERATING SYSTEMS CONSIDERING LOAD BALANCING M. Nkravan and M. H. Kashan Department of Electrcal Computer Islamc Azad Unversty, Shahrar Shahreqods

More information

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

Concurrent Apriori Data Mining Algorithms

Concurrent Apriori Data Mining Algorithms Concurrent Apror Data Mnng Algorthms Vassl Halatchev Department of Electrcal Engneerng and Computer Scence York Unversty, Toronto October 8, 2015 Outlne Why t s mportant Introducton to Assocaton Rule Mnng

More information

Performance Evaluation of Information Retrieval Systems

Performance Evaluation of Information Retrieval Systems Why System Evaluaton? Performance Evaluaton of Informaton Retreval Systems Many sldes n ths secton are adapted from Prof. Joydeep Ghosh (UT ECE) who n turn adapted them from Prof. Dk Lee (Unv. of Scence

More information

A Hybrid Genetic Algorithm for Routing Optimization in IP Networks Utilizing Bandwidth and Delay Metrics

A Hybrid Genetic Algorithm for Routing Optimization in IP Networks Utilizing Bandwidth and Delay Metrics A Hybrd Genetc Algorthm for Routng Optmzaton n IP Networks Utlzng Bandwdth and Delay Metrcs Anton Redl Insttute of Communcaton Networks, Munch Unversty of Technology, Arcsstr. 21, 80290 Munch, Germany

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

Virtual Machine Migration based on Trust Measurement of Computer Node

Virtual Machine Migration based on Trust Measurement of Computer Node Appled Mechancs and Materals Onlne: 2014-04-04 ISSN: 1662-7482, Vols. 536-537, pp 678-682 do:10.4028/www.scentfc.net/amm.536-537.678 2014 Trans Tech Publcatons, Swtzerland Vrtual Machne Mgraton based on

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

More information

Design and Analysis of Algorithms

Design and Analysis of Algorithms Desgn and Analyss of Algorthms Heaps and Heapsort Reference: CLRS Chapter 6 Topcs: Heaps Heapsort Prorty queue Huo Hongwe Recap and overvew The story so far... Inserton sort runnng tme of Θ(n 2 ); sorts

More information

APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT

APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT 3. - 5. 5., Brno, Czech Republc, EU APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT Abstract Josef TOŠENOVSKÝ ) Lenka MONSPORTOVÁ ) Flp TOŠENOVSKÝ

More information

Vectorization of Image Outlines Using Rational Spline and Genetic Algorithm

Vectorization of Image Outlines Using Rational Spline and Genetic Algorithm 01 Internatonal Conference on Image, Vson and Computng (ICIVC 01) IPCSIT vol. 50 (01) (01) IACSIT Press, Sngapore DOI: 10.776/IPCSIT.01.V50.4 Vectorzaton of Image Outlnes Usng Ratonal Splne and Genetc

More information

Greedy Technique - Definition

Greedy Technique - Definition Greedy Technque Greedy Technque - Defnton The greedy method s a general algorthm desgn paradgm, bult on the follong elements: confguratons: dfferent choces, collectons, or values to fnd objectve functon:

More information

Air Transport Demand. Ta-Hui Yang Associate Professor Department of Logistics Management National Kaohsiung First Univ. of Sci. & Tech.

Air Transport Demand. Ta-Hui Yang Associate Professor Department of Logistics Management National Kaohsiung First Univ. of Sci. & Tech. Ar Transport Demand Ta-Hu Yang Assocate Professor Department of Logstcs Management Natonal Kaohsung Frst Unv. of Sc. & Tech. 1 Ar Transport Demand Demand for ar transport between two ctes or two regons

More information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

AN ENHANCED APPROACH TO SURROUND SOUND DECODER DESIGN

AN ENHANCED APPROACH TO SURROUND SOUND DECODER DESIGN AN ENHANCED APPROACH TO SURROUND SOUND DECODER DESIGN J.D.Moore, Dr J.P Wakefeld Unversty of Huddersfeld, Queensgate, Huddersfeld HD1 3DH, UK ABSTRACT Ths paper hghlghts several key mprovements that can

More information

CSCI 104 Sorting Algorithms. Mark Redekopp David Kempe

CSCI 104 Sorting Algorithms. Mark Redekopp David Kempe CSCI 104 Sortng Algorthms Mark Redekopp Davd Kempe Algorthm Effcency SORTING 2 Sortng If we have an unordered lst, sequental search becomes our only choce If we wll perform a lot of searches t may be benefcal

More information

USING GRAPHING SKILLS

USING GRAPHING SKILLS Name: BOLOGY: Date: _ Class: USNG GRAPHNG SKLLS NTRODUCTON: Recorded data can be plotted on a graph. A graph s a pctoral representaton of nformaton recorded n a data table. t s used to show a relatonshp

More information

Machine Learning: Algorithms and Applications

Machine Learning: Algorithms and Applications 14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of

More information

Module Management Tool in Software Development Organizations

Module Management Tool in Software Development Organizations Journal of Computer Scence (5): 8-, 7 ISSN 59-66 7 Scence Publcatons Management Tool n Software Development Organzatons Ahmad A. Al-Rababah and Mohammad A. Al-Rababah Faculty of IT, Al-Ahlyyah Amman Unversty,

More information

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation Internatonal Conference on Logstcs Engneerng, Management and Computer Scence (LEMCS 5) Maxmum Varance Combned wth Adaptve Genetc Algorthm for Infrared Image Segmentaton Huxuan Fu College of Automaton Harbn

More information

A Genetic Algorithm Based Dynamic Load Balancing Scheme for Heterogeneous Distributed Systems

A Genetic Algorithm Based Dynamic Load Balancing Scheme for Heterogeneous Distributed Systems Proceedngs of the Internatonal Conference on Parallel and Dstrbuted Processng Technques and Applcatons, PDPTA 2008, Las Vegas, Nevada, USA, July 14-17, 2008, 2 Volumes. CSREA Press 2008, ISBN 1-60132-084-1

More information

Control strategies for network efficiency and resilience with route choice

Control strategies for network efficiency and resilience with route choice Control strateges for networ effcency and reslence wth route choce Andy Chow Ru Sha Centre for Transport Studes Unversty College London, UK Centralsed strateges UK 1 Centralsed strateges Some effectve

More information

5 The Primal-Dual Method

5 The Primal-Dual Method 5 The Prmal-Dual Method Orgnally desgned as a method for solvng lnear programs, where t reduces weghted optmzaton problems to smpler combnatoral ones, the prmal-dual method (PDM) has receved much attenton

More information

An Image Fusion Approach Based on Segmentation Region

An Image Fusion Approach Based on Segmentation Region Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua

More information

On the Virtues of Parameterized Uniform Crossover

On the Virtues of Parameterized Uniform Crossover On the Vrtues of Parameterzed Unform Crossover Wllam M. Spears Naval Research aboratory Washngton, D.C. 20375 USA spears@ac.nrl.navy.ml Kenneth A. De Jong George Mason Unversty Farfax, VA 22030 USA kdejong@ac.gmu.edu

More information

GENETIC ALGORITHMS APPLIED FOR PATTERN GENERATION FOR DOWNHOLE DYNAMOMETER CARDS

GENETIC ALGORITHMS APPLIED FOR PATTERN GENERATION FOR DOWNHOLE DYNAMOMETER CARDS GENETIC ALGORITHMS APPLIED FOR PATTERN GENERATION FOR DOWNHOLE DYNAMOMETER CARDS L. Schntman 1 ; B.C.Brandao 1 ; H.Lepkson 1 ; J.A.M. Felppe de Souza 2 ; J.F.S.Correa 3 1 Unversdade Federal da Baha- Brazl

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

ELEC 377 Operating Systems. Week 6 Class 3

ELEC 377 Operating Systems. Week 6 Class 3 ELEC 377 Operatng Systems Week 6 Class 3 Last Class Memory Management Memory Pagng Pagng Structure ELEC 377 Operatng Systems Today Pagng Szes Vrtual Memory Concept Demand Pagng ELEC 377 Operatng Systems

More information

Backpropagation: In Search of Performance Parameters

Backpropagation: In Search of Performance Parameters Bacpropagaton: In Search of Performance Parameters ANIL KUMAR ENUMULAPALLY, LINGGUO BU, and KHOSROW KAIKHAH, Ph.D. Computer Scence Department Texas State Unversty-San Marcos San Marcos, TX-78666 USA ae049@txstate.edu,

More information

GSLM Operations Research II Fall 13/14

GSLM Operations Research II Fall 13/14 GSLM 58 Operatons Research II Fall /4 6. Separable Programmng Consder a general NLP mn f(x) s.t. g j (x) b j j =. m. Defnton 6.. The NLP s a separable program f ts objectve functon and all constrants are

More information