Parallel Branch and Bound Algorithm - A comparison between serial, OpenMP and MPI implementations

Size: px
Start display at page:

Download "Parallel Branch and Bound Algorithm - A comparison between serial, OpenMP and MPI implementations"

Transcription

1 Journal of Physcs: Conference Seres Parallel Branch and Bound Algorthm - A comparson between seral, OpenMP and MPI mplementatons To cte ths artcle: Luco Barreto and Mchael Bauer 2010 J. Phys.: Conf. Ser Vew the artcle onlne for updates and enhancements. Related content - Prostate brachytherapy treatmentplan optmzaton W D D'Souza, R R Meyer, B R Thomadsen et al. - Towards a Resource Reservaton Approach for an Opportunstc Computng Envronment Elza Gomes and M A R Dantas - Treatment plannng for brachytherapy: an nteger programmng model, two computatonal approaches and experments wth permanent prostate mplant plannng Eva K Lee, Rchard J Gallagher, Davd Slvern et al. Recent ctatons - Mohand Mezmaz et al Ths content was downloaded from IP address on 16/04/2018 at 13:40

2 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 Parallel Branch and Bound Algorthm - A comparson between seral, OpenMP and MPI mplementatons Luco Barreto and Mchael Bauer Department of Computer Scence Mddlesex College - The Unversty of Western Ontaro London, ON Canada {lbarret6, bauer}@csd.uwo.ca Abstract. Ths paper presents a comparson of an extended verson of the regular Branch and Bound algorthm prevously mplemented n seral wth a new parallel mplementaton, usng both MPI (dstrbuted memory parallel model) and OpenMP (shared memory parallel model). The branch-and-bound algorthm s an enumeratve optmzaton technque, where fndng a soluton to a mxed nteger programmng (MIP) problem s based on the constructon of a tree where nodes represent canddate problems and branches represent the new restrctons to be consdered. Through ths tree all nteger solutons of the feasble regon of the problem are lsted explctly or mplctly ensurng that all the optmal solutons wll be found. A common approach to solve such problems s to convert sub-problems of the mxed nteger problem to lnear programmng problems, thereby elmnatng some of the nteger constrants, and then tryng to solve that problem usng an exstng lnear program approach. The paper descrbes the general branch and bound algorthm used and provdes detals on the mplementaton and the results of the comparson. 1. Introducton Integer Programmng problems (IP) are specal cases of optmzaton problems where the varables can only assume nteger values. Mxed Integer Programmng problems (MIP) are specal cases where only some of the varables are restrcted to nteger values. Optmzaton problems wth nteger varables can also be lnear or nonlnear, dependng on the terms of ther objectve functon and ther constrants. However, n general, the terms IP and MIP are almost always assocated wth problems that have lnear features. The optmzaton algorthms consdered n ths wor focus on mxed nteger programmng, where some varables assume nteger values and whle other varables tae on contnuous values. In consderng only problems wth lnear constrants and lnear objectve functons, there are mportant dfferences between IP and MIP n relaton to Lnear Programmng (LP): LP there are necessary and suffcent condtons of optmalty proved theoretcally that can be used to effcently test whether a gven feasble soluton s an optmal soluton. These condtons are used to develop algebrac methods such as the smplex method. IP/MIP there are no nown optmalty condtons to test whether a gven feasble soluton s optmal. It s necessary to perform the mplct or explct comparson of all feasble solutons of the problem. c 2010 IOP Publshng Ltd 1

3 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 Approaches to solve IP and MIP problems are numerous, usually specalzed n accordance wth a partcular applcaton. We can dvde these approaches nto two broad famles, wth dstnct characterstcs: Heurstc Optmzaton may get good results for problems n whch classcal optmzaton methods mght fal (such as many nteger varables and very complex constrants), but there s no guarantee of optmalty. Examples of such approaches are genetc algorthms [5][8] and tabu search [3][4]. Classcal Optmzaton for convex problems t s guaranteed that the soluton s optmal. In ths famly there are the methods of enumeraton, such as the zero-one enumeraton algorthm [1] and the branch-and-bound algorthm [2], used n ths wor. We focus specfcally on usng a branch-and-bound approach for fndng an optmal soluton to MIP problems, ncludng beng able to fnd all optmal solutons when there are multple ones. The branch-and-bound algorthm s an enumeratve optmzaton technque, where fndng a soluton to an MIP problem s based on the constructon of a tree where nodes represent canddate problems (subproblems) and branches represent the new restrctons to be consdered. Through ths tree, all nteger solutons of the feasble regon of the problem are lsted explctly or mplctly ensurng that all the optmal solutons wll be found. A common approach to solve such problems s to convert the sub-problems that arse n solvng nteger lnear programmng problems to lnear programmng problems, that s, wth no nteger constrants, and to solve those problems usng an exstng lnear programmng solver. The resultng soluton can be used to set bounds on the possble solutons or to help select varables to create restrctons that can be used to create sub-problems. The paper presents a comparson of a branch and bound algorthm mplemented wth three dfferent approaches: seral, shared memory model (OpenMP) and dstrbuted memory model (MPI). The mplemented branch and bound algorthm uses the GNU/GLPK [9] as the Lnear Programmng (LP) solver to fnd optmal solutons to mxed nteger lnear problems. The remnder of the paper s organzed as follows. Secton 2 descrbes some mportant theoretcal aspects of branch-and-bound. Secton 3 descrbes the branch-and-bound algorthm tself. Secton 4 descrbes the computatonal mplementaton. Secton 5 shows a comparson of the approaches. Secton 6 presents conclusons about ths wor and fnally Secton 7 dentfes future wor. 2. Theoretcal Aspects The branch-and-bound algorthm s an enumeratve technque, n whch a soluton s found based on the constructon of a tree n whch nodes represent the problem canddates and branches represent the new restrctons to be consdered. Through ths tree, all nteger solutons of the problem feasble regon are lsted explctly or mplctly ensurng that all the optmal solutons wll be found. The overall structure of the branch-and-bound algorthm has three ey elements, separaton, relaxaton and prunng [2]. Separaton uses the tactc of dvde and conquer n order to solve the problem (P). In order to fnd the soluton of P, t s decomposed nto two or more descendant subproblems, generatng a lst of canddate problems (CP). At a subsequent step n the algorthm, a canddate s selected from the lst of canddate problems and the algorthm tres to solve that problem. If a soluton cannot be found to that problem, that problem s agan decomposed and ts descendants are added to the lst of canddate problems. If the selected problem can be solved, then a new soluton s obtaned. The objectve functon value of ths new soluton s then compared wth the value of the ncumbent soluton, whch s the best feasble soluton nown so far. If the new soluton s better than the ncumbent soluton, t becomes the new ncumbent. Then, the algorthm returns to the lst and selects the next canddate. Ths procedure s repeated untl the lst s empty, and the soluton of the problem s taen as the fnal ncumbent soluton. The usual way to carry out separaton of an nteger programmng problem s through contradctory constrants n a sngle nteger varable (separaton varable or branchng varable). Thus, from the 2

4 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 orgnal problem (called node zero), two new descendant sub-problems are created, whch are easer to solve than the orgnal one, snce a constrant was added to the separaton varable. Each generated node has an assocated canddate sub-problem and each branch ndcates the addton of a constrant related to the varable used n the separaton. Therefore, as the algorthm moves down n the tree, the vable regon of the generated descendants becomes more restrcted. The most common approach to relaxaton s the elmnaton of the ntegral constrants (nteger varables) where the nteger or mxed nteger problem s converted to a general LP. Relaxaton assumes that the orgnal nteger varables may tae fractonal values and then the resultng LP problem s solved. The obtaned optmal soluton usually has several varables wth non-nteger values. Among these varables, one must be selected for separaton. Once the separaton and ncluson of new descendants n canddates lst s completed, the algorthm must select from among the stored canddates the next to be evaluated and, f necessary, separated successvely untl the lnear problem soluton becomes nteger, or unfeasble or worse than the ncumbent, meanng that the canddate sub-problem can be removed from the lst (pruned), not producng any more descendants. Ths procedure repeats untl the lst of canddates sub-problems becomes empty. 3. Branch and bound Algorthm Consder the MIP (P) whose general form s gven by: (P) mnmze subject to T z = c x (1) Ax b x nteger I x j real j J (2) Where c s the vector of costs, x s the vector of varables ntegers (x ) and real (x j ), A s the constrant coeffcent matrx, b s the vector of rght sde of the nequaltes, I s the set of nteger varables and J s the set of contnuous varables. The branch-and-bound algorthm to solve (P) has the followng steps: 1. Start-up: set the number of actve nodes n = 0, set the frst ncumbent and ntalze the lst of canddate sub-problems wth the orgnal problem (P). 2. Convergence test: f the canddate lst s empty then ths means that the process s over and the current ncumbent soluton s the optmal soluton of the problem. Otherwse, contnue. 3. Canddate selecton: among the canddate sub-problems not yet pruned, choose the one that wll be the next to be evaluated and remove t from the lst. Solve the LP problem related to the relaxed CP and store the optmal soluton as a lower bound for all ts descendants, selected problem ( R ) * znf = z ( CP R ). 4. Prunng tests: the sub-problem ( CP ) may be pruned f t meets one of the followng condtons: a) f ( CP R ) has no feasble soluton; * * b) f znf > z, where z s the actual ncumbent value; c) f the optmal soluton of ( CP R ) s nteger and feasble n ( CP ). In ths case, f the optmal value s lower than the ncumbent value, mae * z = znf and apply the prevous test (b) for all canddates sub-problems not yet pruned. CP was pruned, return to the Step 2. If the canddate sub-problem ( ) 3

5 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/ Separaton: from the sub-problem ( CP ), select a varable for separaton from those that are nteger and stll have contnuous value. For the chosen varable x, whose current value s x *, generate two new descendant sub-problems and add them to the canddate lst. The new sub- problems are generated by addng to ( ) CP the followng restrctons: n+ 1 * ( PC ): x x (3) n+ 2 * ( PC ): x x + 1 (4) where * x s the largest nteger not greater than * x. Set n = n + 2 and return to Step Performance Enhancement The algorthm effcency s drectly related to the method of selectng the next canddate subproblem that wll be evaluated (Step 3) and the separaton varable (Step 5). Moreover, the exstence of a good ntal ncumbent ncreases the effectveness of the prunng test (b), thereby reducng the number of canddate sub-problems that need to be evaluated. Although there s no systematc technque to determne whch one of the canddate sub-problems leads more qucly to the soluton, some heurstc rules can be used, such as search by creaton order, by depth, by the relaxed soluton value, by the soluton estmated value, among others. A very common technque used to select the canddate sub-problem s the LIFO rule (Last In, Frst Out), whch produces a depth search. Ths strategy allows that the descendant sub-problem be solved from the prevous problem (because they dffer n only by a sngle varable) and mnmzes the memory requrements to store the canddates nformaton. The way whch these selectons are made drectly nfluences the number of nodes that need to be evaluated, determnng, thus, the computatonal effort. On the other hand, there are methods that use estmates of the objectve functon value to select from among all canddates the most promsng sub-problem. Ths rule mnmzes the total number of problems to be evaluated, but at the same tme can drastcally ncrease the memory requrements, snce the successve LP problems may not have the same smlarty that exsts n the depth search Varable Selecton J When the soluton of the relaxed canddate ( PC R ) has several nteger varables wth contnuous value, one of them must be selected to be separated. An nadequate varable choce mples evaluaton of several descendants sub-problems that could be elmnated by prunng ther predecessor. There s no systematc technque to dentfy the optmal separaton varable, but there are emprcal rules that ndcate whch varables are most attractve. If smplcty s the goal, the separaton varable can be selected from a predetermned sequence usng the objectve functon coeffcents, hghest to lowest cost, for example, or beng more specfc dependng on the problem s features. A more effectve alternatve s based on the search for the varable that has the hghest value n terms of the estmated ncrease of the objectve functon, whch can be obtaned from one of the followng technques: [MAX;MAX]: always choose the varable whch causes the greatest degradaton n the objectve functon n order to qucly obtan a descendant sub-problem that mght be pruned. For the canddate sub-problem node, the selecton method for the separaton varable j follows the expresson below: j + j max{ max( P f, P (1 f ))} 4

6 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 [MAX;MIN]: choose the varable whch the mnmum varaton causes the maxmum mpact. Ths ensures that both descendants contrbute to prunng the sub-problem. For the canddate sub-problem node, the selecton method for the separaton varable j follows the expresson below: j + j max{ mn( P f, P (1 f ))} [MIN;MAX]: choose the varable whch the maxmum varaton causes the mnmum mpact. Smlarly, both descendants contrbute to prunng ths sub-problem. For the canddate sub-problem node, the selecton method for the separaton varable j follows the expresson below: j + j mn{ max( P f, P (1 f ))} [MIN;MIN]: always choose the varable whch causes the lowest degradaton n the objectve functon n order to qucly obtan a vable soluton. Ths s the most conservatve method. For the canddate sub-problem node, the selecton method for the separaton varable j follows the expresson below: j + j mn{ mn( P f, P (1 f ))} Another ssue nvolved n varable selecton s the possblty of separatng a varable whch was assgned an nteger value. It often occurs n a branch-and-bound executon that a specfc varable gets an nteger value prematurely,.e., wthout any lmts to mpose that nteger value. In ths case, accordng to Step 5 of the branch-and-bound algorthm, ths varable would not be selected for separaton (because t s nteger), and f all others varables were also ntegers, the soluton would be consdered vable, not generatng descendants. Ths would generate a fal n order to guarantee the reachng of all alternatve solutons to the problem. By separatng the varables wth nteger values whch stll have a lower bound mnor than ts upper bound, the algorthm guarantees that all nteger varables nvolved n the problem wll be evaluated n an mplct or explct way, exhaustng all possble separaton opportuntes and allowng that all alternatve solutons to the problem may be found Multple Optmal Solutons Some problems may have multple optmal nteger solutons, such as optmzaton models that descrbe the power networs expanson problem for both transmsson [] and dstrbuton [6][7]. In ths case, obtanng all the solutons s very mportant. Although the objectve functon value s the same for all solutons, one soluton may be greater than the other when tang nto account other factors that were not explctly ncluded n the optmzaton model. If these other factors have some relevance, one may also want to retan sub-optmal solutons. Ths can be easly accomplshed when solutons found are stored n an ordered lst, as used n ths wor. To ensure that all solutons are found, t s necessary to mae changes n the prunng (Step 4) and separaton (Step 5) processes. In Step 4 (c), one soluton s consdered nteger when all the nteger varables cannot assume other values,.e., when they all have nequalty constrants wth upper and lower bounds both ntegers and dentcal. Thus, the Step 4 (c) of the conventonal algorthm should be replaced by: c) f the optmal soluton of ( ) CP and all nteger varables n the optmal CP R s feasble n ( ) lb ub lb ub soluton of ( CP R ) have no degree of freedom.e. x = x, I, where x and x are the lower and the upper bound vectors of nteger varables of canddate problem ( CP R ), respectvely. In ths case, f the optmal value s lower than the ncumbent value, mae 5

7 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 * z = znf and apply the prevous test (b) for all canddates sub-problems not pruned yet. Otherwse, f the optmal value s equal to the ncumbent value, store the new alternatve soluton. Notce that the components of vectors x lb and x ub are always nteger values generated by the separaton process, accordng to equatons (3) and (4). In Step 5, n addton to usng the varables that have contnuous values, t s also necessary to consder separatng the nteger varables that have some lb ub degree of freedom,.e., wth x < x. 4. Computatonal Implementaton The branch-and-bound algorthm was mplemented n C++ and was adapted to three dfferent versons: one seral and two parallel. The two parallel mplementatons respectvely used OpenMP [11] and MPI [10]. The parallel versons are based on the seral approach, just mang use of the necessary parallel calls. However, all approaches use the same programmng logc. In the system, several parameters may be confgured, such as: maxmum sze of the branch-and-bound tree; dfferent tolerances for nteger and equalty; selecton crtera of the separaton varable; crtera to select a node; mnmum depth to open the tree 1 ; occupaton percentage of the avalable space n the tree to exchange sort order of the actve nodes n order to preserve avalable memory; possblty of separaton n nteger varables. The central data structure used n the branch-and-bound algorthm stores all the nformaton that characterzes each canddate sub-problem (actve node). The nformaton assocated wth each node of the soluton tree s dvded nto two groups: actve nodes nformaton related to the canddate sub-problems; tree nodes nformaton of the predecessors nodes that were separated (usng a contradctory constrant) generatng the actve nodes. In ths wor, each node stores only the nformaton relatng to the last separaton; the rest of the nformaton s assocated wth the tree nodes. Ths procedure avods data redundancy. The canddate sub-problems, when generated, are nserted nto the lsts of actve nodes; several lsts are mantaned dependng on a partcular sort crteron. The avalable orderng possbltes on actve nodes are as follows: node number (LIFO, Last In Frst Out); node depth; node relaxed soluton (LP value); node estmated soluton (usng pseudo-costs); node estmated soluton weghted wth depth (deepest nodes are more mportant). 1 Before adoptng the chosen strategy for selectng the canddate node, the program wll expand the tree breadth-frst, untl all actve nodes have the same depth as the mnmum chosen n the confguraton. 6

8 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 Snce all lsts are crcularly-lned, the algorthm can select the node that has any of the extreme values (mnmum or maxmum, n each of the crtera) drectly. Moreover, as all lsts are smultaneously updated, t s possble to swtch between any crtera durng the soluton process wthout the need for reorderng, whch have hgh computatonal cost. Whenever a canddate node s pruned (satsfyng some of the crtera presented n Step 4 of the algorthm n Secton 3) t s removed from the lst of actve nodes, together wth all ts predecessors that do not possess actve descendants, recursvely, from the node that was pruned towards the source node of the tree. As an example, consder the problem P: mn v = 5n + 2n13 + 2n23 + β s. a. 350n + 400n13 + β n + 210n23 + β 200 ( P) β 0 nj 0 j n, n13 e n23 ntegers The branch-and-bound algorthm for ths problem can be represented by Fgure 1, where each node s represented for a crcle wth a number that ndcates ts generaton s order (ndex of the algorthm). In the sde of each node there s the relaxed soluton and f t that soluton s nteger or worse than the ncumbent, t s related to a pruned node, represented by a colourful crcle. Fgure 1. Branch-and-bound ntermedate tree. 7

9 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 The constrants that are added for each separaton are shown nsde the rectangles n the respectve branch. Thus, for any node, the assocated problem can be determned gong bac on the tree towards the root node (node 0) and addng the correspondent constrants for that path. For example, the canddate problem represented by node 7 s the orgnal problem (P) wth the addton of the followng constrants: n n The tree starts wth Node 0 through the lnear problem resoluton (P R ), whch s obtaned from the relaxaton of the nteger constrants of the expresson 1. ( P ) R mn s. a. v = 5n 350n 350n β 0 n j + 2n + 400n + 210n 0 j n β + β β 200 The lnear problem soluton presents two varables wth contnuous values: n 4 = 7 and n 13 = Thus, t does not satsfy the prunng crteron and the process contnues wth the separaton of the problem. The frst non-nteger varable ( n ) s chosen, as the separaton varable. In ths case, the two descendants nodes are generated by the constrants n 0 (Node 1) and n 1 (Node 2). In ths part of the process there are two canddates that may be chosen to be evaluated, represented by Nodes 1 and 2. Supposng that the selectng crteron s to evaluate the last generated node. So, Node 2 wll be the chosen one. The equvalent relaxed problem soluton relaxed problem wth the constrant n 1 stll has a varable wth contnuous value ( n 13 = 0. 5 ). Then, the problem of the Node 2 should be separated and, n ths case, only one varable can be selected, ( n 13 ). After the generaton of the Node 2 descendants (Nodes 3 and 4), Node 4 s selected and evaluated. Thus, the frst nteger soluton for the orgnal problem s obtaned and ths soluton becomes the * 4 4 ncumbent ( v = v = vnf = 7 ). As the ncumbent value was changed, the remaned canddates (Nodes 1 and 3) are evaluated, but t s not possble to mae any prunng. Agan, the last generated canddate s selected (Node 3) and ts varable n s separated. Both descendants of the Node 3 (Nodes 5 and 6) have nteger solutons that are dscarded snce they are worse than the exstng ncumbent ( v = vnf = 55 and v = vnf = 10 ). The algorthm now goes bac to ts canddates lst and selects the only avalable node (Node 1) whch s separated n ts varable n 23 generatng Nodes 7 and 8. Tang the last one to be evaluated, * 8 8 we obtaned an nteger soluton better than the ncumbent, whch s updated to v = v = vnf = 4. Next, Node 7 s evaluated and dscarded (wth ts descendants) snce ts lower bound s greater than 7 * the ncumbent ( v nf = 201 > v ). Nodes 9 and 10 are represented only to complete the enumeraton. In the way that the process evolved, t was necessary to solve 9 lnear problems to guarantee the enumeraton of all possble alternatves. The optmal soluton for the orgnal problem was obtaned after 8 LPs: 8

10 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 v * * n * 13 n n * 23 = 4 = 0 = 1 = 1 An mportant ssue regardng performance n enumeraton algorthms s related to how good the ntal ncumbent soluton s. A good ntal soluton sgnfcantly ncreases the effcency of the prunng test (4b) and so, reduces the number of canddates that need to be evaluated. For the example of Fgure 1, an ntal ncumbent v * = 5, would be enough to mmedately prune Node 2 (and all ts descendants), reducng then the total number of evaluatons. Consderng the data structures used and the crtera for node orderng and selecton of the separaton varable, the seral branch-and-bound algorthm was mplemented accordng to the flowcharts of Fgures 2 and 3. A summary of the parallel mplementaton can be seen n Fgure Parallel Implementaton MPI In the MPI mplementaton, the master node ntalzes the program, creates a base tree wth a lst of actve nodes and sends those actve nodes to the slaves whenever they need a problem to wor on. The master also controls the process, ncludng lst of solutons, statstcs, etc. Whle the master s creatng a base tree, the slaves also create a copy of the same base tree. Each slave receves an ndex for a node from the master, then ntalzes ts own tree and starts the B&B. All the calculaton s done by the slaves. Once a soluton s found, t s sent to the master who updates all slaves. The master node wats for any message from the slaves and t does no other calculatons durng the B&B process. Whenever a slave fnshes ts job, t sends a message to the master asng for a new node and process contnues. As the slaves wor n dfferent nodes wth ther own tree, there s no possblty of conflct n accessng the same part of the memory OPENMP In the OpenMP mplementaton, there s no concept of master and slaves. The cpu0 node s the one responsble for ntalzng the program and creatng a common tree wth a lst of actve nodes. In ths mplementaton, the nodes have a flag ndcatng whether or not they are beng used. After ths creaton, all nodes start the B&B requestng the frst avalable node from the lst of actve nodes. So, each slave wors n dfferent part of the same tree. Of course, as the memory s shared, once somethng happens, the varables are automatcally updated. Wth both mplementatons there are possbltes for further mprovements and those wll be dscussed n the Sectons 6 and 7. 9

11 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 Fgure 2. Seral branch-and-bound algorthm Part 1 Fgure 3. Seral branch-and-bound algorthm Part 2. Fgure 4. Summary of the parallel branch-and-bound algorthm In the next secton we present results comparng the three mplementatons for solvng the same MIP. 10

12 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/ Results The ntal comparson of the algorthms to date made use of an optmzaton problem prevously presented [6][7]. The problem s a smplfed three-phase electrcal dstrbuton networ consstng of 18 nodes (2 substatons and 16 nodes wth loads) and 24 branches operatng under 13800V. The topology of ths networ s shown n Fgure 5 n whch rectangles denote the substatons and crcles are the nodes where loads are concentrated. Branches drawn as contnuous lnes denote the ntal networ (those wth a sngle lne are part of the fxed networ and those wth double lnes are canddates for replacement), and branches drawn as dashed lnes are canddates for addton (and are 118 not part of the ntal networ). The search space has approxmately 2 combnatons. The results are llustrated n Fgure 6. The problem s fully detaled n the above reference. Fgure 5: Dagram of the 18-node networ. [7] Fgure 6: 18-node networ - Soluton wth multstage expanson. [7] 11

13 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 The results for the tests were carred out on two dfferent systems: - HP Lnux cluster runnng XC 3.1 wth 267 nodes, 4 cores each, connected va Myrnet 2g (narwhal.sharcnet.ca) for the MPI and Seral Implementaton. - SUSE Enterprse 10 Lnux Cluster runnng wth 8 nodes, 1 core per socet, NUMA (sly.sharcnet.ca) for the OpenMP. Even though these are dfferent systems wth dfferent archtectures, the goal of the experment was to compare the wall tme of the parallel mplementatons wth the seral one to support the dea that classc optmzaton methods can stll be used n modern dstrbuted systems for many dfferent areas. Table 2: Results for the Electrcal Dstrbuton Networ Seral Approach #LPs Wall Tme (s) Table 3: Results for the Electrcal Dstrbuton Networ Parallel Approaches #Proc #LPs Wall Tme (s) Speedup MPI OpenMP MPI OpenMP MPI OpenMP x 1.1x x 1.2x x 1.3x x 1.9x x 2.1x Table 2 summarzes the executon of the soluton to the problem for the seral mplementaton. The results show the requred number of lnear programmng problems (sub-problems) (#LPs) and the total tme requred n seconds. The tme was measured by the algorthm and corresponds to the total wall cloc tme. Table 3 summarzes the executon for both parallel mplementatons: MPI and OpenMP. The results show the number of processors used (#Proc), the requred number of lnear programmng problems (sub-problems) (#LPs) for each approach related to #Proc, as well as the total tme spent across all processors and the calculated speedup. For the tests we used the [MAX;MAX] separaton varable selecton strategy. The dfference between the number of LPs n the parallel executons s not related to a dfferent strategy, but s due to several other factors. In the MPI mplementaton, as the memory s dstrbuted and the processors have to exchange messages, each node s creatng ts own tree and because of that t could tae more tme to fnd a good ncumbent soluton. Many unnecessary nodes are evaluated before fndng a good one, snce the algorthm has to ensure the enumeraton of all possble solutons wthout prunng anythng that could be promsng. On the other hand, n the OpenMP approach, fewer nodes were evaluated because the algorthm always found good solutons earler than the MPI (for ths partcular problem). Also, the OpenMP mplementaton s more smlar to the seral one. Unfortunately, other ssues wth the OpenMP mplementaton were found and are reported n the conclusons. 6. Conclusons Based on the ntal expermental results some conclusons can be drawn: The seral mplementaton wors reasonably well.

14 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 The OpenMP mplementaton always generated fewer sub-problems than the seral, but the speedup was not as good as the MPI mplementaton. Unfortunately, the chosen LP solver (GLPK) [9] was not thread safe, whch made t mpossble to run a shared memory model approach wthout the use of some explct mechansms of control n the mplementaton. Ths drastcally decreased the performance. Also, the use of a flag n order to control the lst of actve nodes to avod race condton has a cost that decreases performance, as well. We plan to address those problems n future wor. The MPI mplementaton seemed to provde a more robust way to splt the branch and bound search. Though t generated more sub-problems for some cases, t too the least tme n all executons and always ran perfectly. It s mportant to menton here that before the parallel computaton starts, the algorthm prepares a base tree for the MPI mplementaton and the nodes are sent to the slaves whenever they need. Ths base tree s ntally statc, whch means that there are a specfc number of nodes avalable to be evaluated. When ths number s reached, no more nodes are avalable to be sent and the allocated processors become dle, watng the others to fnsh the job. Ths s a problem that wll be addressed n future wor. As ndcated, these results are based on our partcular problem and, whle complex, addtonal experments need to be run to better compare the algorthms and approaches to parallel mplementaton as well as to help mprove the mplementaton. Our ntal mpresson s that the MPI approach wors reasonable well and wth much less headache than the OpenMP one. 7. Future Wor Our next steps nclude fndng a better way to evaluate nodes n parallel and fndng a better way to eep all processors worng full tme. We also want to test the algorthm usng others LP solvers that are thread-safe and faster. Fnally, we ntend to compare the mplementatons to several other problems ncludng some from other domans, testng the dfferent separaton varable choce that we had mplemented as well. References [1] Balas, E. (1965). An addtve algorthm for solvng lnear programs wth zero-one varables, Operatons Research, Vol. 13, No. 4, pp [2] Geoffron, A.M. and R.E. Marsten (1972). Integer Programmng Algorthms: A Framewor and State-of-the-Art Survey, Management Scence, Vol. 18, No. 9, pp [3] Glover, F. (1989). Tabu Search Part I, ORSA Journal on Computng, Vol. 1, No. 3, pp [4] Glover, F. (1990). Tabu Search Part II, ORSA Journal on Computng, Vol. 2, No. 1, pp [5] Goldberg, D.E. (1989). Genetc Algorthms n Search, Optmzaton and Machne Learnng, Addson-Wesley Professonal, 432 p. [6] Haffner, S., L.F. Perera, L.A. Perera and L. Barreto (2008a). Multstage model for dstrbuton expanson plannng wth dstrbuted generaton - Part I: problem formulaton. IEEE Transactons on Power Delvery, Vol. 23, No. 2, pp [7] Haffner, S., L.F. Perera, L.A. Perera and L. Barreto (2008b). Multstage model for dstrbuton expanson plannng wth dstrbuted generaton - Part II: numercal results. IEEE Transactons on Power Delvery, Vol. 23, No. 2, pp [8] Holland, J.H. (1992). Adaptaton n Natural and Artfcal Systems: An Introductory Analyss wth Applcatons to Bology, Control, and Artfcal Intellgence, The MIT Press, 228 p. [9] Mahorn, A., (2001). GLPK Lnear Programmng Kt Manual GLPK documentaton, 13

15 Hgh Performance Computng Symposum (HPCS2010) Journal of Physcs: Conference Seres 256 (2010) 0018 IOP Publshng do: / /256/1/0018 Moscow Avaton Insttute, Moscow, Russa, February 2001, avalable n [10] MPI-forum - [11] OpenMP - [] Romero, R., Montcell, A., Garca, A. e Haffner, S. (2002). Test systems and mathematcal models for transmsson networ expanson plannng, IEE Proc.-Gener. Transm Dstrb., Vol. 1491, No. 1, pp

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

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

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

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

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

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

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

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

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

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

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

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

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

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

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

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

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

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

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

CMPS 10 Introduction to Computer Science Lecture Notes

CMPS 10 Introduction to Computer Science Lecture Notes CPS 0 Introducton to Computer Scence Lecture Notes Chapter : Algorthm Desgn How should we present algorthms? Natural languages lke Englsh, Spansh, or French whch are rch n nterpretaton and meanng are not

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

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal

More information

Chapter 6 Programmng the fnte element method Inow turn to the man subject of ths book: The mplementaton of the fnte element algorthm n computer programs. In order to make my dscusson as straghtforward

More information

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and

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

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

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

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

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

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

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

More information

Repeater Insertion for Two-Terminal Nets in Three-Dimensional Integrated Circuits

Repeater Insertion for Two-Terminal Nets in Three-Dimensional Integrated Circuits Repeater Inserton for Two-Termnal Nets n Three-Dmensonal Integrated Crcuts Hu Xu, Vasls F. Pavlds, and Govann De Mchel LSI - EPFL, CH-5, Swtzerland, {hu.xu,vasleos.pavlds,govann.demchel}@epfl.ch Abstract.

More information

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following.

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following. Complex Numbers The last topc n ths secton s not really related to most of what we ve done n ths chapter, although t s somewhat related to the radcals secton as we wll see. We also won t need the materal

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

Intra-Parametric Analysis of a Fuzzy MOLP

Intra-Parametric Analysis of a Fuzzy MOLP Intra-Parametrc Analyss of a Fuzzy MOLP a MIAO-LING WANG a Department of Industral Engneerng and Management a Mnghsn Insttute of Technology and Hsnchu Tawan, ROC b HSIAO-FAN WANG b Insttute of Industral

More information

A Facet Generation Procedure. for solving 0/1 integer programs

A Facet Generation Procedure. for solving 0/1 integer programs A Facet Generaton Procedure for solvng 0/ nteger programs by Gyana R. Parja IBM Corporaton, Poughkeepse, NY 260 Radu Gaddov Emery Worldwde Arlnes, Vandala, Oho 45377 and Wlbert E. Wlhelm Teas A&M Unversty,

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

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

SENSITIVITY ANALYSIS IN LINEAR PROGRAMMING USING A CALCULATOR

SENSITIVITY ANALYSIS IN LINEAR PROGRAMMING USING A CALCULATOR SENSITIVITY ANALYSIS IN LINEAR PROGRAMMING USING A CALCULATOR Judth Aronow Rchard Jarvnen Independent Consultant Dept of Math/Stat 559 Frost Wnona State Unversty Beaumont, TX 7776 Wnona, MN 55987 aronowju@hal.lamar.edu

More information

LECTURE NOTES Duality Theory, Sensitivity Analysis, and Parametric Programming

LECTURE NOTES Duality Theory, Sensitivity Analysis, and Parametric Programming CEE 60 Davd Rosenberg p. LECTURE NOTES Dualty Theory, Senstvty Analyss, and Parametrc Programmng Learnng Objectves. Revew the prmal LP model formulaton 2. Formulate the Dual Problem of an LP problem (TUES)

More information

Improving Low Density Parity Check Codes Over the Erasure Channel. The Nelder Mead Downhill Simplex Method. Scott Stransky

Improving Low Density Parity Check Codes Over the Erasure Channel. The Nelder Mead Downhill Simplex Method. Scott Stransky Improvng Low Densty Party Check Codes Over the Erasure Channel The Nelder Mead Downhll Smplex Method Scott Stransky Programmng n conjuncton wth: Bors Cukalovc 18.413 Fnal Project Sprng 2004 Page 1 Abstract

More information

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints

TPL-Aware Displacement-driven Detailed Placement Refinement with Coloring Constraints TPL-ware Dsplacement-drven Detaled Placement Refnement wth Colorng Constrants Tao Ln Iowa State Unversty tln@astate.edu Chrs Chu Iowa State Unversty cnchu@astate.edu BSTRCT To mnmze the effect of process

More information

UB at GeoCLEF Department of Geography Abstract

UB at GeoCLEF Department of Geography   Abstract UB at GeoCLEF 2006 Mguel E. Ruz (1), Stuart Shapro (2), June Abbas (1), Slva B. Southwck (1) and Davd Mark (3) State Unversty of New York at Buffalo (1) Department of Lbrary and Informaton Studes (2) Department

More information

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

Simulation Based Analysis of FAST TCP using OMNET++

Simulation Based Analysis of FAST TCP using OMNET++ Smulaton Based Analyss of FAST TCP usng OMNET++ Umar ul Hassan 04030038@lums.edu.pk Md Term Report CS678 Topcs n Internet Research Sprng, 2006 Introducton Internet traffc s doublng roughly every 3 months

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

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

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

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

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

Review of approximation techniques

Review of approximation techniques CHAPTER 2 Revew of appromaton technques 2. Introducton Optmzaton problems n engneerng desgn are characterzed by the followng assocated features: the objectve functon and constrants are mplct functons evaluated

More information

Optimization Methods: Integer Programming Integer Linear Programming 1. Module 7 Lecture Notes 1. Integer Linear Programming

Optimization Methods: Integer Programming Integer Linear Programming 1. Module 7 Lecture Notes 1. Integer Linear Programming Optzaton Methods: Integer Prograng Integer Lnear Prograng Module Lecture Notes Integer Lnear Prograng Introducton In all the prevous lectures n lnear prograng dscussed so far, the desgn varables consdered

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

Comparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments

Comparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments Comparson of Heurstcs for Schedulng Independent Tasks on Heterogeneous Dstrbuted Envronments Hesam Izakan¹, Ath Abraham², Senor Member, IEEE, Václav Snášel³ ¹ Islamc Azad Unversty, Ramsar Branch, Ramsar,

More information

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

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

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

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

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

Convolutional interleaver for unequal error protection of turbo codes

Convolutional interleaver for unequal error protection of turbo codes Convolutonal nterleaver for unequal error protecton of turbo codes Sna Vaf, Tadeusz Wysock, Ian Burnett Unversty of Wollongong, SW 2522, Australa E-mal:{sv39,wysock,an_burnett}@uow.edu.au Abstract: Ths

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

Hybrid Heuristics for the Maximum Diversity Problem

Hybrid Heuristics for the Maximum Diversity Problem Hybrd Heurstcs for the Maxmum Dversty Problem MICAEL GALLEGO Departamento de Informátca, Estadístca y Telemátca, Unversdad Rey Juan Carlos, Span. Mcael.Gallego@urjc.es ABRAHAM DUARTE Departamento de Informátca,

More information

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science

EECS 730 Introduction to Bioinformatics Sequence Alignment. Luke Huan Electrical Engineering and Computer Science EECS 730 Introducton to Bonformatcs Sequence Algnment Luke Huan Electrcal Engneerng and Computer Scence http://people.eecs.ku.edu/~huan/ HMM Π s a set of states Transton Probabltes a kl Pr( l 1 k Probablty

More information

Available online at Available online at Advanced in Control Engineering and Information Science

Available online at   Available online at   Advanced in Control Engineering and Information Science Avalable onlne at wwwscencedrectcom Avalable onlne at wwwscencedrectcom Proceda Proceda Engneerng Engneerng 00 (2011) 15000 000 (2011) 1642 1646 Proceda Engneerng wwwelsevercom/locate/proceda Advanced

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

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

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

Learning to Project in Multi-Objective Binary Linear Programming

Learning to Project in Multi-Objective Binary Linear Programming Learnng to Project n Mult-Objectve Bnary Lnear Programmng Alvaro Serra-Altamranda Department of Industral and Management System Engneerng, Unversty of South Florda, Tampa, FL, 33620 USA, amserra@mal.usf.edu,

More information

An efficient iterative source routing algorithm

An efficient iterative source routing algorithm An effcent teratve source routng algorthm Gang Cheng Ye Tan Nrwan Ansar Advanced Networng Lab Department of Electrcal Computer Engneerng New Jersey Insttute of Technology Newar NJ 7 {gc yt Ansar}@ntedu

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

Brave New World Pseudocode Reference

Brave New World Pseudocode Reference Brave New World Pseudocode Reference Pseudocode s a way to descrbe how to accomplsh tasks usng basc steps lke those a computer mght perform. In ths week s lab, you'll see how a form of pseudocode can be

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

Research of Dynamic Access to Cloud Database Based on Improved Pheromone Algorithm

Research of Dynamic Access to Cloud Database Based on Improved Pheromone Algorithm , pp.197-202 http://dx.do.org/10.14257/dta.2016.9.5.20 Research of Dynamc Access to Cloud Database Based on Improved Pheromone Algorthm Yongqang L 1 and Jn Pan 2 1 (Software Technology Vocatonal College,

More information

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

More information

AMath 483/583 Lecture 21 May 13, Notes: Notes: Jacobi iteration. Notes: Jacobi with OpenMP coarse grain

AMath 483/583 Lecture 21 May 13, Notes: Notes: Jacobi iteration. Notes: Jacobi with OpenMP coarse grain AMath 483/583 Lecture 21 May 13, 2011 Today: OpenMP and MPI versons of Jacob teraton Gauss-Sedel and SOR teratve methods Next week: More MPI Debuggng and totalvew GPU computng Read: Class notes and references

More information

Hierarchical clustering for gene expression data analysis

Hierarchical clustering for gene expression data analysis Herarchcal clusterng for gene expresson data analyss Gorgo Valentn e-mal: valentn@ds.unm.t Clusterng of Mcroarray Data. Clusterng of gene expresson profles (rows) => dscovery of co-regulated and functonally

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

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on

More information

CE 221 Data Structures and Algorithms

CE 221 Data Structures and Algorithms CE 1 ata Structures and Algorthms Chapter 4: Trees BST Text: Read Wess, 4.3 Izmr Unversty of Economcs 1 The Search Tree AT Bnary Search Trees An mportant applcaton of bnary trees s n searchng. Let us assume

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

Conditional Speculative Decimal Addition*

Conditional Speculative Decimal Addition* Condtonal Speculatve Decmal Addton Alvaro Vazquez and Elsardo Antelo Dep. of Electronc and Computer Engneerng Unv. of Santago de Compostela, Span Ths work was supported n part by Xunta de Galca under grant

More information

11. APPROXIMATION ALGORITHMS

11. APPROXIMATION ALGORITHMS Copng wth NP-completeness 11. APPROXIMATION ALGORITHMS load balancng center selecton prcng method: vertex cover LP roundng: vertex cover generalzed load balancng knapsack problem Q. Suppose I need to solve

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

The Shortest Path of Touring Lines given in the Plane

The Shortest Path of Touring Lines given in the Plane Send Orders for Reprnts to reprnts@benthamscence.ae 262 The Open Cybernetcs & Systemcs Journal, 2015, 9, 262-267 The Shortest Path of Tourng Lnes gven n the Plane Open Access Ljuan Wang 1,2, Dandan He

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

Configuration Management in Multi-Context Reconfigurable Systems for Simultaneous Performance and Power Optimizations*

Configuration Management in Multi-Context Reconfigurable Systems for Simultaneous Performance and Power Optimizations* Confguraton Management n Mult-Context Reconfgurable Systems for Smultaneous Performance and Power Optmzatons* Rafael Maestre, Mlagros Fernandez Departamento de Arqutectura de Computadores y Automátca Unversdad

More information

Related-Mode Attacks on CTR Encryption Mode

Related-Mode Attacks on CTR Encryption Mode Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory

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

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

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

EVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS

EVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS Academc Research Internatonal ISS-L: 3-9553, ISS: 3-9944 Vol., o. 3, May 0 EVALUATIO OF THE PERFORMACES OF ARTIFICIAL BEE COLOY AD IVASIVE WEED OPTIMIZATIO ALGORITHMS O THE MODIFIED BECHMARK FUCTIOS Dlay

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

Parallel Numerics. 1 Preconditioning & Iterative Solvers (From 2016)

Parallel Numerics. 1 Preconditioning & Iterative Solvers (From 2016) Technsche Unverstät München WSe 6/7 Insttut für Informatk Prof. Dr. Thomas Huckle Dpl.-Math. Benjamn Uekermann Parallel Numercs Exercse : Prevous Exam Questons Precondtonng & Iteratve Solvers (From 6)

More information

Adaptive Weighted Sum Method for Bi-objective Optimization

Adaptive Weighted Sum Method for Bi-objective Optimization 45th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamcs & Materals Conference 19-22 Aprl 2004, Palm Sprngs, Calforna AIAA 2004-1680 Adaptve Weghted Sum Method for B-objectve Optmzaton Olver de Weck

More information

A Modelling and a New Hybrid MILP/CP Decomposition Method for Parallel Continuous Galvanizing Line Scheduling Problem

A Modelling and a New Hybrid MILP/CP Decomposition Method for Parallel Continuous Galvanizing Line Scheduling Problem ISIJ Internatonal, Vol. 58 (2018), ISIJ Internatonal, No. 10 Vol. 58 (2018), No. 10, pp. 1820 1827 A Modellng and a New Hybrd MILP/CP Decomposton Method for Parallel Contnuous Galvanzng Lne Schedulng Problem

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

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

More information

A fast algorithm for color image segmentation

A fast algorithm for color image segmentation Unersty of Wollongong Research Onlne Faculty of Informatcs - Papers (Arche) Faculty of Engneerng and Informaton Scences 006 A fast algorthm for color mage segmentaton L. Dong Unersty of Wollongong, lju@uow.edu.au

More information

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide Lobachevsky State Unversty of Nzhn Novgorod Polyhedron Quck Start Gude Nzhn Novgorod 2016 Contents Specfcaton of Polyhedron software... 3 Theoretcal background... 4 1. Interface of Polyhedron... 6 1.1.

More information