Globalized Nelder Mead method for engineering optimization

Size: px
Start display at page:

Download "Globalized Nelder Mead method for engineering optimization"

Transcription

1 Computers and Structures xxx (4) xxx xxx Globalzed Nelder Mead method f engneerng optmzaton Marco A. Luersen a,b, *, Rodolphe Le Rche c a CNRS UMR 638/LMR, INSA de Rouen, Lab. de Mecanque, Avenue de lõunverste, St Etenne du Rouvray 768, France b Mechancal Department, CEFET-PR, Av. e de embro, Curtba, 83F9, Brazl c CNRS UMR 546/SMS, Ecole des Mnes de Sant Etenne, 58 Cours Faurel, 43, Sant Etenne, France Receved December ; accepted 4 March 4 Abstract One of the fundamental dffcultes n engneerng desgn s the multplcty of local solutons. Ths has trggered great effts to develop global search algthms. Globalty, however, often has a prohbtvely hgh numercal cost f real problems. A fxed cost local search, whch sequentally becomes global s developed. Globalzaton s acheved by probablstc restart. A spatal probablty of startng a local search s bult based on past searches. An mproved Nelder Mead algthm makes the local optmzer. It accounts f varable bounds. It s addtonally made me robust by rentalzng degenerated smplexes. The resultng method, called Globalzed Bounded Nelder Mead (GBNM) algthm, s partcularly adapted to tackle multmodal, dscontnuous optmzaton problems, f whch t s uncertan that a global optmzaton can be affded. Dfferent strateges f restartng the local search are dscussed. Numercal experments are gven on analytcal test functons and composte lamnate desgn problems. The GBNM method compares favably to an evolutonary algthm, both n terms of numercal cost and accuracy. Ó 4 Cvl-Comp Ltd. and Elsever Ltd. All rghts reserved. Keywds: Global optmzaton; Nelder Mead; Composte lamnate desgn. Introducton Complex engneerng optmzaton problems are characterzed by calculaton ntensve system smulatons, dffcultes n estmatng senstvtes (when they exst), the exstence of desgn constrants, and a multplcty of local solutons. Acknowledgng the last pont, much research has been devoted to global optmzaton (e.g., [,]). The * Crespondng auth. Address: CNRS UMR 638/LMR, INSA de Rouen, Lab. de Mecanque, Avenue de lõunverste, St Etenne du Rouvray 768, France. Tel.: ; fax: E-mal address: (M.A. Luersen). hgh numercal cost of global optmzers has been at the gn of subsequent effts to speed up the search ether by addng problem specfc knowledge to the search, by mxng effcent, local algthms wth global algthms. There are many ways n whch local and global searches can cooperate. The smplest strategy s to lnk the searches n seres, meanng that, frstly, a global optmzaton of lmted cost s executed, the soluton of whch s refned by a local search. An example of the seral hybrd s gven n [3] where smulated annealng, the global optmzer, s coupled wth a sequental quadratc programmng and a Nelder Mead algthm. A large number of parallel local-global searches have been proposed [,4,5] and analyzed [6,7]. In these cases, teratons of global and local algthms are ntertwned /$ - see front matter Ó 4 Cvl-Comp Ltd. and Elsever Ltd. All rghts reserved. do:.6/j.compstruc.4.3.7

2 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx One can further classfy parallel hybrds nto those where the local searches converge, and those where local searches may be prematurely stopped. Memetc genetc algthms [8] and multstart methods (e.g., determnstc restart n [9], random restarts n []) are examples of the fmer. The latter are usually based on clusterng steps, where local searches approachng already expled regons of the desgn space are abandoned [6,]. When consderng a real engneerng optmzaton problem, a common stuaton s that the affdable total number of analyses s lmted, that the presence of spurous local mnma s unknown, and that t s uncertan f t wll be possble to complete as few as two local searches. Nevertheless, achevng global results remans an objectve of the optmzer. Ths typcally occurs when dealng wth an unknown functon of less than varables, f whch one s wllng to wat f about evaluatons of the objectve functon. In such a case, a local-global method based on restart s the safest strategy because t can termnate n a sht tme (the length of a sngle local search). The method descrbed n ths artcle, the Globalzed Bounded Nelder Mead algthm (GBNM) s meant to be a black-box local-global approach to real optmzaton problems. A restart procedure that uses an adaptve probablty densty keeps a memy of past local searches. Lmts on varables are taken nto account through projecton. Fnally, GBNM can be appled to dscontnuous (no gradent nfmaton needed), non-convex functons, snce the local searches are based on a varant of the Nelder Mead algthm []. Improvements to the Nelder Mead algthm consst of smplex degeneracy detecton and handlng through rentalzaton. Ths paper s structured as follows. The GBNM algthm s descrbed n Secton, and Secton 3 repts numercal experments on analytcal functons and composte lamnated plate desgn problems. In partcular, dfferent strateges f restartng the mproved Nelder Mead search are numercally dscussed. The GBNM algthm s also compared to a steady-state evolutonary algthm [].. Globalzaton of a local search by probablstc restart Local optmzers can make up a global search when repeatedly started from dfferent ponts. The smplest restart methods ntalze the search ether from a regular grd of ponts, from randomly chosen ponts. In the frst case, one needs to know how many restarts wll be perfmed to calculate the sze of the mesh. In the other case, knowledge of past searches s not used, so that the same local optma may be found several tmes, costng vast unnecessary efft. In the current wk, the number of restarts s unknown befehand because a maxmum number of analyses s mposed and the cost of each local search s unknown. A grd method cannot be appled here. Also, a memy of prevous local searches s kept by buldng a spacal probablty densty of startng a search... Probablstc restart The probablty, p(x), of havng sampled a pont x s descrbed here by a Gaussan Parzen-wndows approach [3]. Ths method can be consdered as a smoothed verson of the hstograms technques, the hstograms beng centered at selected sampled ponts. The probablty p(x) s wrtten, pðxþ ¼ N X N ¼ p ðxþ; ðþ where N s the number of ponts already sampled, and p s the nmal multdmensonal probablty densty functon, p ðxþ ¼ ðpþ n ðdetðrþþ exp ðx x Þ T R ðx x Þ ; ðþ n s the dmenson (number of varables) and R the covarance matrx, R ¼ 6 4 r... r n : ð3þ The varances, r j, are estmated by the relaton, ; r j ¼ a xmax j x mn j ð4þ where a s a postve parameter that controls the length of the Gaussans, and x max j and x mn j are the bounds n the jth drecton. te that, n der to keep the method as smple and cost effectve as possble, the varances are kept constant. Ths strategy would have a cost n terms of total number of analyses. The probablty densty s such that R pðxþdx ¼, but snce a bounded doman X s consdered, a bounded probablty ~pðxþ s ntroduced, ~pðxþ ¼ pðxþ Z M ; M ¼ pðxþ dx; ð5þ X so that R ~pðxþdx ¼. X The probablty densty of samplng a new pont, /(x), s a probablty densty of not havng sampled x befe. F ts estmaton we adopt the followng assumpton: only the hghest pont x H of ~pðxþ has a null probablty of beng sampled at the next teraton. So, the probablty /(x) s calculated as,

3 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx 3 H p( xh) p p( x ) Xmn H p( x ) H ~pðxþ /ðxþ ¼R ; H ¼ max ðh ~pðxþþdx ~pðxþ: ð6þ xx X Fg. llustrates p(x), ~pðxþ and H ~pðxþ, n a undmensonal doman. The maxmzaton of / s not perfmed exactly, frstly because of ts numercal cost, and secondly, as wll be seen n Secton 3., because t would be detrmental to the search. Instead, N r ponts are chosen randomly and the pont that maxmzes / s selected to ntate the next search. te that, n der to maxmze /, t s necessary to calculate nether M (5) n H (6): the maxmum of / s the mnmum of p, sop only s calculated. Three parameters nfluence the probablty densty p and, consequently, the startng ponts: the ponts that are kept f the probablty calculaton, p; the number of random ponts used to maxmze /, N r ; and the Gaussans length parameter, a. Ther settng s dscussed n the numercal results (Secton 3.). The probablstc restart procedure can be appled to any local optmzer. In ths case, an mproved Nelder Mead algthm s proposed... An mproved Nelder Mead search x H Xmax Fg.. Probablty densty functons. Ω The gnal Nelder Mead algthm [] and the strategy f boundng varables are summarzed n Appendx A. The GBNM algthm dffers from the Nelder Mead method because of a set of restart optons. The purpose of the restarts s twofold. Frstly, probablstc restarts based on the densty p (Eq. ()) am at repeatng local searches untl a fxed total cost, C max has been reached. The probablty of havng located a global optmum ncreases wth the number of probablstc restarts. Ths s the globalzed aspect of the method. In the current mplementaton of probablstc restart, the sze of the new smplex, a (defned n Eq. (A.)), s a unfm random varable taken between % and % of the smallest doman dmenson. Secondly, restarts are used to check and mprove convergence of the algthm. The two-restart schemes that are convergence related ntalze a new smplex from the current best vertex. The small and large test restarts use a small and large smplex of szes a s and a l, respectvely (see Eq. (A.)). Convergence of the local Nelder Mead searches s estmated through three crtera, the small, flat degenerate smplex tests. The smplex s small f! max k¼;...;nþ X n k¼ x k x max x mn < e s ; ð7þ where x k s the th component of the kth edge, x mn and x max are the bounds n the th drecton, and e s s a termnaton tolerance. The smplex s flat f jf H f L j < e s ; ð8þ where f H and f L are the hghest and lowest objectve functons n the smplex, and e s s a tolerance value. The smplex s degenerated f t has collapsed nto a subspace of the search doman. Ths s the most common symptom of a faled Nelder Mead search [4] because the method cannot escape the subspace. Me precsely, a smplex s called degenerated here f t s nether small, n touches a varable bound, and one of the two followng condtons s satsfed: mn k¼;n kek k max k¼;n kek k < e s3 det½eš Q ke k k < e s4; k ð9þ where e k s the kth edge, e s the edge matrx, kæk represents the Eucldean nm, and e s3 and e s4 are small postve constants. The lnkng of the three restarts and three convergence tests n the GBNM algthm s shown n Fg.. A memy of past convergence locatons s kept, thus preventng unnecessarly spendng computatons on already analyzed ponts (thrd test, T3, n the flow chart of Fg. ). When the smplex s flat, a probablstc restart s perfmed (T4). A smplex whch s degenerated nduces a large test teraton (T8). When the optmalty of the convergence pont s unsure, such as a convergence on a varable bound where the smplex has degenerated (T6), a small test, that stands f an optmalty check, s perfmed. If the small smplex returns to the same convergence pont, t s consdered to be a local optmum. It should be remembered that the Kuhn and Tucker condtons of mathematcal programmng are not applcable to the present non-dfferentable framewk. The tolerances f small and degenerated smplces, e s and [e s3,e s4 ], respectvely, may be dffcult to tune, so that a smplex whch s becomng small may be tagged as degenerated befe. Thus, f a degeneraton s detected twce consecutvely at the same pont, the pont s taken as a possble optmum, and a probablstc restart s called. Smlarly, f a degeneraton s detected after a small test, ths pont s also saved as a possble optmum, and a large test s dered.

4 4 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx INITIALIZATION restart = PROBABILISTIC Save ntal pont To calculate the probablty densty Nelder-Mead teraton wth projecton (cf. Fg. 5) Save ntal pont RESTART To calculate the probablty densty PROBABILISTIC LARGE TEST SMALL TEST T. of evaluatons > Cmax T Converge The smplex s: small flat degenerated Save the smplex best pont END restart = PROBABILISTIC T3 Pont already known as a local optmum restart = PROBABILISTIC T4 Smplex flat T5 restart = PROBABILISTIC Save local optmum (SMALL LARGE TEST and return to the same pont) (not SMALL TEST and pont not on the bounds and small smplex) T6 restart = SMALL TEST Possble local optmum LARGE TEST PROBABILISTIC and not return to the same pont and pont on the bounds restart = LARGE TEST Save the smplex best pont (possble local optmum) T7 SMALL TEST and not return to the same pont and not on the bounds T8 restart = LARGE TEST t SMALL TEST and ponts not on the bounds and smplex not small and smplex degenerated Degeneracy case restart = SMALL TEST Fg.. Restarts and convergence tests lnkng n GBNM. Once the GBNM algthm termnates, the lst of possble local (eventually global) optma makes the results of the search. In practce, the calculaton of many local global optma s a beneft of the method n comparson wth global optmzers that provde a sngle soluton (e.g., evolutonary algthms). Fnally, t

5 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx 5 should be noted that, n der to fnd all the local optma, the number of restarts should be larger equal to the number of local optma. Typcally, me restars are requred because searches started at dfferent ponts may converge to the same local soluton. 3. Numercal results In Secton 3., the choce of GBNM parameters s dscussed. Results on an analytcal functon are gven n Secton 3.and composte lamnate desgn problems are addressed n Sectons 3.3 and 3.4. The GBNM method s compared to an evolutonary algthm (EA). The evolutonary algthm [] has a steady-state structure [5] wth real encodng, contnuous crossover, and Gaussan mutaton of varance r mut ¼ðx max x mn Þ =6. F far comparsons, the parameters of the EA chosen f each test are the ones that perfm best n ndependent trals among all combnatons of populaton szes ( 5), mutaton probabltes (.5.) and crossover probabltes (.4.5). 3.. GBNM parameters choce 3... The Gaussans length parameter, a In ths wk, a s set to., whch means that one standard devaton away from the Gaussan mean pont covers about % of the doman Ponts kept f the probablty calculaton Three strateges have been compared n terms of probablty of not fndng at least one local mnmum, P nfm : the x Õs used n Eqs. () and () are () the startng ponts, () the startng and local convergence ponts, () all the ponts sampled durng the search. One should remember that local convergence ponts are never duplcated (test T3 on Fg. ). From prelmnary tests, t has been observed that strategy () s memy and tme consumng, wth degraded perfmance wth respect to strateges () and (). F these reasons, t has not been consdered f further testng. Strateges () and () are tested wth N r varyng from to, on three functons, 8 f ðx ; x Þ¼ þ :ðx x Þ þð x Þ þ ð x Þ þ snð:5x Þ snð:7x x Þ x ; x ½; 5Š; >< f ðx ; x Þ¼ 4 :x þ 3 x4 x þ x x þð 4 þ 4x Þx x ; x ½ 3; 3Š; f 3 ðx ; x Þ¼ x 5: x 4p þ 5 x p 6 þ cosðx Þþ 8p >: x ½ 5; Š; x ½; 5Š; ðþ f has 4 local mnma, f 6, and f 3 3 (see Fg. 3). f s known as the Sx Humps Camel Back functon and f 3 as the BrannÕs rcos functon. F the frst and the second strateges, results after 5 analyss and based on ndependents runs are presented n Tables and, respectvely. The second strategy perfms best, ndependently of N r. It shows that the startng and local convergence ponts effcently summarze the topology of the basns of attracton. Ths scheme s chosen to update p Number of random ponts, N r If N r s equal to, the rentalzaton s random. If N r s large, the ntal ponts are dstrbuted based on the ntal and convergence ponts of past searches, whch nduces a regular grd-lke pattern. tng N r to a small number, larger than, gves a based-random rentalzaton. It should be seen as a compromse between the grd and the random strateges. Optmum value of N r depends on the test functon: f the basns of attracton are regularly dstrbuted, restarts followng a regular pattern (.e., N r large) are optmal, and vce versa. From the tests results on the three multmodal functons presented n Table, the optmal strategy s N r large f f and f 3, whle t s N r =f f. N r = s chosen as a compromse f general functon optmzaton. 3.. GrewankÕs functon mnmzaton Consder the mnmzaton of the GrewankÕs test functon, F ðx ;...; x n Þ¼ X n x Yn x cos pff ; 4n ¼ ¼ x ½ ; Š; ðþ where the dmenson s n=and the global mnmum s. at x =., =,n. Ths functon has many local mnma. Fg. 4 shows t n the one-dmensonal case (n=), x[,]. Table 3 compares GBNM and the best pont n the populaton of an evolutonary algthm (EA) at,, 5 and, functon calls. Table 3 averages ndependent runs where the startng pont of the GBNM s randomly selected. The followng crteron s used to consder the EA has convergedd to the global mnmum: n k^x x k < ; ðþ where ^x s the best pont found, and x * the global mnmum. The man observaton s that the GBNM method fnds, on average, better objectve functon values, wth a hgher probablty of fndng the global mnmum than the EA does. The advantage of the GBNM method s substantal at a low number of analyses, and slowly decreases as numercal cost grows.

6 6 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx 5 Functon f 5 Functon f3 4 3 x x (a) x -5 5 (c) x.5 Functon f.5 x (b) x Fg. 3. Test functons: contour plots Bucklng load maxmzaton Composte lamnates are made of stacked layers where each layer has ented fbers melted n an sotropc matrx (see sketch n Fg. 5). The desgn problems addressed here am at fndng the optmal entaton of the fbers wthn each layer, h, where h s a contnuous varable bounded by and 9. The plates are analyzed Table Probablty P nfm of not fndng at least one of the local mnma (C max =5 analyses, runs, only the startng ponts are kept f the probablty densty calculaton) N r f f f 3 (random restart) Table Probablty P nfm of not fndng at least one of the local mnma (C max =5 analyses, runs, startng and convergence ponts are kept f the probablty densty calculaton) N r f f f 3 (random restart)

7 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx 7 F(x) x Fg. 4. GrewankÕs test functon (n=). usng the classcal lamnaton they and an elastc lnear bucklng model (see [6]). Consder a smply suppted carbon-epoxy square plate, subjected to n-plane compressve loads N x =N y, as shown n Fg. 5. The plate s balanced and symmetrc and has 3layers, each of whch are.5 mm thck. The elastc materal propertes of the layers are E =5 GPa, E =5 GPa, G =5 GPa, m =.35. The lamnate s desgned to maxmze ts bucklng load (assocated to the most crtcal bucklng mode). Snce the plate s balanced and symmetrc, there are eght contnuous desgn varables, the ply entatons, whch are bounded between and 9. Ths problem has a unque mnmum, all ples ented at 45. te also that the outermost ples have me nfluence on the bucklng behav than the nnermost ples. Table 4 compares GBNM and the evolutonary algthm (EA), showng the stackng sequences found, after 3, 5 and analyses based on ndependents runs. At 3 evaluatons, the bucklng load of the desgns proposed by GBNM and EA are equvalent. One notces that the frst local search of GBNM has not always converged at that pont. A steady dfference between GBNM and EA desgns at 3 analyses can be seen: GBNM, whch s by nature a me ented search, converges faster on the me senstve outerples than EA does, and vce versa on the nnerples. From 5 evaluatons on, GBNM converges me accurately to the optmum than EA does Composte plate longtudnal stffness maxmzaton A 6-ply balanced and symmetrc plate, made of glass-epoxy, s to be desgned by maxmzng the longtudnal stffness E x (see [7, p. 47]). The elastc propretes f the glass-epoxy layers are E =45 GPa, E = GPa, Table 3 Comparson of GBNM (N r =) and the best of an EA populaton on GrewankÕs functon, runs (average ± standard devaton) analyses analyses 5 analyses, analyses Mnmum functon value Probablty of fndng the global mnmum Mnmum functon value Probablty of fndng the global mnmum Mnmum functon value Probablty of fndng the global mnmum Mnmum functon value Probablty of fndng the global mnmum GBNM 9.3±6.79 /.56±.499 /.947±.74 5/.98±.4 3/ EA ±5.537 / 4.86±.9 /.9±.96 /.57±. 9/ θ = fber entaton N y N x a a Fg. 5. Smply suppted rectangular plate subjected to n plane loads.

8 8 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx Table 4 Bucklng load maxmzaton, runs, N r = (stackng sequence: average ± standard devaton) 3 analyses GBNM [±45.4 ±44.97 ±45.±45.38 ±45.38 ±44.97 ±43.8 ±49.65] s Std. Dev. ±.47 ±.54 ±.83 ±4.6 ±4.37 ±.53 ±7.46 ±3.47 EA [±45.9 ±44.9 ±45.3 ±44.55 ±44.78 ±45. ±45.6 ±44.85] s Std. Dev. ±.75 ±.96 ±.67 ±3. ±3.66 ±5.9 ±8.8 ± analyses GBNM [±45. ±45.3 ±45. ±45.8 ±45.4 ±44.97 ±45.5 ±45.8] s Std. Dev. ±.7 ±. ±.39 ±.4 ±.9 ±.46 ±.9 ±4. EA [±45.3 ±44.95 ±44.99 ±44.95 ±44.8 ±45. ±44.7 ±46.45] s Std. Dev. ±.9 ±.8 ±.6 ±.9 ±. ±3. ±4.49 ±.3 analyses GBNM [±45. ±45. ±45. ±45. ±44.99 ±45. ±44.98 ±45.] s Std. Dev. ±.±.±.3 ±.5 ±.4 ±.6 ±.5 ±.44 EA [±44.96 ±44.98 ±44.96 ±45.7 ±44.99 ±44.9±45.3 ±45.7] s Std. Dev. ±.6 ±.6 ±.7 ±.95 ±.±.7 ±.7 ±4.95 G =4.5 GPa and m =.3. The plate s balanced and symmetrc, so there are four fber entatons to be found. Ths problem presents 6 local mnma whch are all the combnatons of the and 9 entatons. The global maxmum has all ples ented at. Ths example shows that local solutons exst f smple composte lamnate desgn problems. Table 5 gves the average number of local optma found after analyses based on GBNM runs as a functon of N r. The frst startng pont s chosen randomly. One observes that the average number of optma found grows wth N r. Ths s expected snce the optma are regularly dstrbuted n the doman. Meover, wthn a budget of functon evaluatons, the global optmum s always found because ts basn of attracton s the largest one. Table 5 GBNM E x maxmzaton, analyses, runs N r Average number of optma found Standard devaton Acknowledgment The frst auth would lke to express hs thanks to the Brazlan fundng agency CNPq f the fnancal suppt durng ths research. 4. Concludng remarks A local/global optmzaton method based on probablstc restart has been presented. Local searches are perfmed by an mproved Nelder Mead algthm where desgn varables can be bounded and some search falure cases prevented. The method, called Globalzed and Bounded Nelder Mead search, does not need senstvtes and constructvely uses computer resources up to a gven lmt. It yelds a lst of canddate local optma, whch contan, wth an ncreasng probablty n terms of computer tme, global solutons. The GBNM method s smple n ts prncples, and the afementonned features make t partcularly useful n an engneerng desgn context. It has been found to compare favably to an evolutonary optmzaton algthm n terms of convergence speed, accuracy of results and ablty to fnd local optma. Appendx A. A Nelder Mead algthm wth bounded varables The Nelder Mead method [] s the most popular drect search method f mnmzng unconstraned real functons. It s based on the comparson of functon values at the n+ vertces x of a smplex. A smplex of sze a s ntalzed at x based on the rule (see [7]), x ¼ x þ pe þ Xn qe k ; ¼ ; n; ða:þ k¼ k6¼ where e are the unt base vects and p ¼ a p n ffffff pffffffffffffffffffffff n þ þ n ; q ¼ a p n ffff pffffffffffffffffffffff n þ : ða:þ

9 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx 9 Smplex ntalzaton P : smplex pont f : objectve functon value at P Ph : smplex pont where the objectve functon assumes ts hghest value Ps : smplex pont where the objectve functon assumes ts second hghest value Determne Ph, Ps, Pl, Pm fh, fs, fl Pl : smplex pont where the objectve functon assumes ts lowest value Pm : centrod smplex pont (not consderng Ph) r : reflecton coeffcent =. Reflecton: Pr = Pm + r (Pm-Ph) If Pr s out of the doman: Projecton on bounds β: contracton coeffcent = / γ : expanson coeffcent = fr < fl fr <= fs fr < fh new Nelder Mead teraton Expanson: Pe = Pm + γ (Pr-Pm) If Pe s out of the doman: Projecton on bounds fe < fr Replace Ph by Pr Contracton : Pc = Pm + β(ph-pm) fc > fh Replace Ph by Pe Replace Ph by Pr Replace all P s by (P + Pl)/ Replace Ph by Pc Converged end Fg. 6. Nelder Mead algthm wth bounded varables. The smplex vertces are changed through reflecton, expanson and contracton operatons n der to fnd an mprovng pont (see Fg. 6). The algthm termnates when the vertces functon values become smlar, whch s measured wth the nequalty, vffffffffffffffffffffffffffffffffffffffffffffffffffff ux t nþ ðf f Þ < e; X nþ f ¼ f ; ða:3þ n n þ ¼ where e s a small postve scalar. The cumulatve effect of the operatons on the smplex s, roughly speakng, to stretch the shape along the descent drectons, and to zoom around local optma. Two comments on the propertes of the algthm are added. Frstly, the Nelder Mead algthm may fal to converge to a local optmum, whch happens n partcular when the smplex collapses nto a subspace. Secondly, the method may escape a regon that would be a basn of attracton f a pontwse descent search f the smplex s large enough. ¼ Ultmately, as the sze of the smplex decreases, the algthm becomes local. The gnal Nelder Mead algthm was conceved f unbounded doman problems. Wth bounded varables, the ponts can leave the doman after ether the reflecton the expanson operaton. It s straghtfward to account f varables bounds by projecton, ( f ðx < x mn Þ; x ¼ x mn ; ða:4þ f ðx > x max Þ; x ¼ x max : The flowchart of the Nelder Mead method shown n Fg. 6 dffers from the gnal method only n the ntalzaton (Eq. (A.)) and n the bounded varables. An mptant sde effect of accountng f the bounded varables through projecton s that t tends to make the smplex collapse nto the subspace of the saturated varables. A specfc convergence test, based on a small smplex rentalzaton at the pont of convergence s then requred (see small test n Secton.).

10 M.A. Luersen, R. Le Rche / Computers and Structures xxx (4) xxx xxx References [] Törn AA, Zlnskas A. Global optmzaton. Berln: Sprnger-Verlag; 989. [] Bäck T. Evolutonary algthms n they and practce. Oxfd Unversty Press; 996. [3] Shang Y, Wan Y, Fromherz MPJ, Crawfd L. Toward adaptve cooperaton between global and local solvers f contnuous constrant problems. n: CPÕ Wkshop on cooperatve Solvers n Constrants Programmng, Pahos, Cyprus, December. [4] Durand N, Allot J-M. A Combned Nelder Mead smplex and genetc algthm. Avalable from: [5] Okamoto M, naka T, Ocha S, Tomnaga D. nlnear numercal optmzaton wth use of a hybrd genetc algthm ncpatng the modfed Powell method. Appl Math Comput 998;9:63 7. [6] Törn AA. A search-clusterng approach to global optmzaton. Towards Global Optm 978;:49 6. [7] Goldberg DE, Voessner S. Optmzng global local search hybrds. In: GECCO 99 Genetc and Evolutonary Computaton Conference, Orlando, FL, USA, 999. p. 8. [8] Moscato P. On Evoluton, search, optmzaton, genetc algthms and martal arts: towards memetc algthms. Caltech Concurrent Computaton Program, C3P Rept 86, 989. [9] Barhen J, Protopopescu V, Rester D. TRUST: A determnstc algthm f global constraned optmzaton. Scence 997;76(May):94 7. [] Hu X, Shonkwller R, Sprull MC. Random restarts n global optmzaton. Techncal Rept, School of Mathematcs, Gega Insttute of Technology, Atlanta, January 994. [] Hckernell FJ, Yuan Y-X. A smple multstart algthm f global optmzaton. OR Trans 997;:. [] Nelder JA, Mead R. A smplex f functon mnmzaton. Comput J 965;7:38 3. [3] Duda OR, Hart PE, Stk DG. Pattern classfcaton. nd ed.. New Yk: John Wley & Sons;. [4] Wrght MH. Drect search methods: one scned, now respectable. In: Dundee Bennal Conference n Numercal Analyss, Harlow, UK, 996. p [5] Syswerda G. A study of reproducton n generatonal and steady state genetc algthms. In: Rawlns GJE, edt. Foundatons of genetc algthms. San Mateo, CA: Mgan Kaufmann; 99. [6] Berthelot J-M. Composte materals mechancal behav and structural analyss. Mechancal engneerng seres. - Sprnger; 999. [7] Haftka RT, Gürdal Z. Elements of structural optmzaton. 3rd rev. and expanded ed.. Kluwer Academc Publshers; 99.

"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *

Research Note APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES * Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789-794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

5.2 Least-Squares Fit to a Straight Line

5.2 Least-Squares Fit to a Straight Line 5. Least-Squares Ft to a Straght Lne total probablty and ch-square mnmzng ch-square for a straght lne solvng the determnants example least-squares ft weghted least-squares wth an example least-squares

More information

CS 2750 Machine Learning. Lecture 17a. Clustering. CS 2750 Machine Learning. Clustering

CS 2750 Machine Learning. Lecture 17a. Clustering. CS 2750 Machine Learning. Clustering Lecture 7a Clusterng Mlos Hauskrecht mlos@cs.ptt.edu 539 Sennott Square Clusterng Groups together smlar nstances n the data sample Basc clusterng problem: dstrbute data nto k dfferent groups such that

More information

A Computer Technique for Solving LP Problems with Bounded Variables

A Computer Technique for Solving LP Problems with Bounded Variables Dhaka Unv. J. Sc. 60(2): 163-168, 2012 (July) A Computer Technque for Solvng LP Problems wth Bounded Varables S. M. Atqur Rahman Chowdhury * and Sanwar Uddn Ahmad Department of Mathematcs; Unversty of

More information

ECON 546: Themes in Econometrics Lab. Exercise 2

ECON 546: Themes in Econometrics Lab. Exercise 2 Unversty of Vctora Department of Economcs ECON 546: Themes n Econometrcs Lab. Exercse Introducton The purpose of ths lab. exercse s to show you how to use EVews to estmate the parameters of a regresson

More information

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by 6 CHAPTER 8 COMPLEX VECTOR SPACES 5. Fnd the kernel of the lnear transformaton gven n Exercse 5. In Exercses 55 and 56, fnd the mage of v, for the ndcated composton, where and are gven by the followng

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

Communication Networks II Contents

Communication Networks II Contents 8 / 1 -- Communcaton Networs II (Görg) -- www.comnets.un-bremen.de Communcaton Networs II Contents 1 Fundamentals of probablty theory 2 Traffc n communcaton networs 3 Stochastc & Marovan Processes (SP

More information

Realization of uniform approximation by applying mean-square approximation

Realization of uniform approximation by applying mean-square approximation Computer Applcatons n Electrcal Engneerng Realzaton of unform approxmaton by applyng mean-square approxmaton Jan Purczyńsk West Pomeranan Unversty of Technology 71-126 Szczecn, ul 26 Kwetna 1, e-mal: janpurczynsk@pspl

More information

II. PROBABILITY OF AN EVENT

II. PROBABILITY OF AN EVENT II. PROBABILITY OF AN EVENT As ndcated above, probablty s a quantfcaton, or a mathematcal model, of a random experment. Ths quantfcaton s a measure of the lkelhood that a gven event wll occur when the

More information

Automatic Robotic arm calibration using parameter optimization techniques

Automatic Robotic arm calibration using parameter optimization techniques Automatc Robotc arm calbraton usng parameter optmzaton technques Project by: Chetan Kalyan Introducton: Robotc manpulaton tasks need accurate calbraton between sensors (eyes) and arms. Standard manual

More information

Luby s Alg. for Maximal Independent Sets using Pairwise Independence

Luby s Alg. for Maximal Independent Sets using Pairwise Independence Lecture Notes for Randomzed Algorthms Luby s Alg. for Maxmal Independent Sets usng Parwse Independence Last Updated by Erc Vgoda on February, 006 8. Maxmal Independent Sets For a graph G = (V, E), an ndependent

More information

Global Optimization Algorithms with Application to Non-Life Insurance

Global Optimization Algorithms with Application to Non-Life Insurance Global Optmzaton Algorthms wth Applcaton to Non-Lfe Insurance Problems Ralf Kellner Workng Paper Char for Insurance Economcs Fredrch-Alexander-Unversty of Erlangen-Nürnberg Verson: June 202 GLOBAL OPTIMIZATION

More information

Solution of Algebraic and Transcendental Equations

Solution of Algebraic and Transcendental Equations CHAPTER Soluton of Algerac and Transcendental Equatons. INTRODUCTION One of the most common prolem encountered n engneerng analyss s that gven a functon f (, fnd the values of for whch f ( = 0. The soluton

More information

Mooring Pattern Optimization using Genetic Algorithms

Mooring Pattern Optimization using Genetic Algorithms 6th World Congresses of Structural and Multdscplnary Optmzaton Ro de Janero, 30 May - 03 June 005, Brazl Moorng Pattern Optmzaton usng Genetc Algorthms Alonso J. Juvnao Carbono, Ivan F. M. Menezes Luz

More information

Nonlinear data mapping by neural networks

Nonlinear data mapping by neural networks Nonlnear data mappng by neural networks R.P.W. Dun Delft Unversty of Technology, Netherlands Abstract A revew s gven of the use of neural networks for nonlnear mappng of hgh dmensonal data on lower dmensonal

More information

L10: Linear discriminants analysis

L10: Linear discriminants analysis L0: Lnear dscrmnants analyss Lnear dscrmnant analyss, two classes Lnear dscrmnant analyss, C classes LDA vs. PCA Lmtatons of LDA Varants of LDA Other dmensonalty reducton methods CSCE 666 Pattern Analyss

More information

SCHEDULING OF CONSTRUCTION PROJECTS BY MEANS OF EVOLUTIONARY ALGORITHMS

SCHEDULING OF CONSTRUCTION PROJECTS BY MEANS OF EVOLUTIONARY ALGORITHMS SCHEDULING OF CONSTRUCTION PROJECTS BY MEANS OF EVOLUTIONARY ALGORITHMS Magdalena Rogalska 1, Wocech Bożeko 2,Zdzsław Heduck 3, 1 Lubln Unversty of Technology, 2- Lubln, Nadbystrzycka 4., Poland. E-mal:rogalska@akropols.pol.lubln.pl

More information

1 Approximation Algorithms

1 Approximation Algorithms CME 305: Dscrete Mathematcs and Algorthms 1 Approxmaton Algorthms In lght of the apparent ntractablty of the problems we beleve not to le n P, t makes sense to pursue deas other than complete solutons

More information

9.1 The Cumulative Sum Control Chart

9.1 The Cumulative Sum Control Chart Learnng Objectves 9.1 The Cumulatve Sum Control Chart 9.1.1 Basc Prncples: Cusum Control Chart for Montorng the Process Mean If s the target for the process mean, then the cumulatve sum control chart s

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

Calculation of Sampling Weights

Calculation of Sampling Weights Perre Foy Statstcs Canada 4 Calculaton of Samplng Weghts 4.1 OVERVIEW The basc sample desgn used n TIMSS Populatons 1 and 2 was a two-stage stratfed cluster desgn. 1 The frst stage conssted of a sample

More information

Section 13.5 Test of Hypothesis for Population Proportion

Section 13.5 Test of Hypothesis for Population Proportion Secton 135 Test of Hypothess for Populaton Proporton As dscussed n Secton 114-115, nference about the probablty of success n ndependent Bernoull trals s same as the mean of ndependent observatons on bnary

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

State function: eigenfunctions of hermitian operators-> normalization, orthogonality completeness

State function: eigenfunctions of hermitian operators-> normalization, orthogonality completeness Schroednger equaton Basc postulates of quantum mechancs. Operators: Hermtan operators, commutators State functon: egenfunctons of hermtan operators-> normalzaton, orthogonalty completeness egenvalues and

More information

Multivariate EWMA Control Chart

Multivariate EWMA Control Chart Multvarate EWMA Control Chart Summary The Multvarate EWMA Control Chart procedure creates control charts for two or more numerc varables. Examnng the varables n a multvarate sense s extremely mportant

More information

I. SCOPE, APPLICABILITY AND PARAMETERS Scope

I. SCOPE, APPLICABILITY AND PARAMETERS Scope D Executve Board Annex 9 Page A/R ethodologcal Tool alculaton of the number of sample plots for measurements wthn A/R D project actvtes (Verson 0) I. SOPE, PIABIITY AD PARAETERS Scope. Ths tool s applcable

More information

S. Malasri, D.A.Halijan and M.L.Keough Department of Civil Engineering Christian Brothers University Memphis, TN 38104. Abstract

S. Malasri, D.A.Halijan and M.L.Keough Department of Civil Engineering Christian Brothers University Memphis, TN 38104. Abstract S. Malasr, D.A.Haljan and M.L.Keough Department of Cvl Engneerng Chrstan Brothers Unversty Memphs, TN 38104 Abstract Ths paper demonstrates an applcaton of the natural selecton process to the desgn of

More information

38123 Povo Trento (Italy), Via Sommarive 14 GENETICALLY-DESIGNED ARBITRARY LENGTH ALMOST DIFFERENCE SETS

38123 Povo Trento (Italy), Via Sommarive 14  GENETICALLY-DESIGNED ARBITRARY LENGTH ALMOST DIFFERENCE SETS UNIVERSITY OF TRENTO DIPARTIMENTO DI INGEGNERIA E SCIENZA DELL INFORMAZIONE 38123 Povo Trento (Italy), Va Sommarve 14 http://www.ds.untn.t GENETICALLY-DESIGNED ARBITRARY LENGTH ALMOST DIFFERENCE SETS G.

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

Facility location I. Chapter 10. Facility location Continuous facility location models

Facility location I. Chapter 10. Facility location Continuous facility location models Faclty locaton I. Chapter 10 Faclty locaton Contnuous faclty locaton models Sngle faclty mnsum locaton problem Sngle faclty mnmax locaton problem Faclty locaton Factors that nfluence the faclty locaton

More information

Optimal portfolios using Linear Programming models

Optimal portfolios using Linear Programming models Optmal portfolos usng Lnear Programmng models Chrstos Papahrstodoulou Mälardalen Unversty, Västerås, Sweden Abstract The classcal Quadratc Programmng formulaton of the well known portfolo selecton problem,

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

Graph Theory and Cayley s Formula

Graph Theory and Cayley s Formula Graph Theory and Cayley s Formula Chad Casarotto August 10, 2006 Contents 1 Introducton 1 2 Bascs and Defntons 1 Cayley s Formula 4 4 Prüfer Encodng A Forest of Trees 7 1 Introducton In ths paper, I wll

More information

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification Lecture 4: More classfers and classes C4B Machne Learnng Hlary 20 A. Zsserman Logstc regresson Loss functons revsted Adaboost Loss functons revsted Optmzaton Multple class classfcaton Logstc Regresson

More information

MODULE 11a Topics: The vector norm of a matrix

MODULE 11a Topics: The vector norm of a matrix MODULE 11a Topcs: The vector norm of a matrx Let denote a norm on R m and R n. Typcally, we thnk of x = x = max x, but t can be any norm. We defne the vector norm of a matrx A by A = max x =1 Ax. We say

More information

Extending Probabilistic Dynamic Epistemic Logic

Extending Probabilistic Dynamic Epistemic Logic Extendng Probablstc Dynamc Epstemc Logc Joshua Sack May 29, 2008 Probablty Space Defnton A probablty space s a tuple (S, A, µ), where 1 S s a set called the sample space. 2 A P(S) s a σ-algebra: a set

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6 PAR TESTS If a WEIGHT varable s specfed, t s used to replcate a case as many tmes as ndcated by the weght value rounded to the nearest nteger. If the workspace requrements are exceeded and samplng has

More information

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12 14 The Ch-squared dstrbuton PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 1 If a normal varable X, havng mean µ and varance σ, s standardsed, the new varable Z has a mean 0 and varance 1. When ths standardsed

More information

CS838-1 Advanced NLP: Conditional Random Fields

CS838-1 Advanced NLP: Conditional Random Fields CS838-1 Advanced NLP: Condtonal Random Felds Xaojn Zhu 2007 Send comments to jerryzhu@cs.wsc.edu 1 Informaton Extracton Current NLP technques cannot fully understand general natural language artcles. However,

More information

Optimal Bidding Strategies for Generation Companies in a Day-Ahead Electricity Market with Risk Management Taken into Account

Optimal Bidding Strategies for Generation Companies in a Day-Ahead Electricity Market with Risk Management Taken into Account Amercan J. of Engneerng and Appled Scences (): 8-6, 009 ISSN 94-700 009 Scence Publcatons Optmal Bddng Strateges for Generaton Companes n a Day-Ahead Electrcty Market wth Rsk Management Taken nto Account

More information

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS 21 22 September 2007, BULGARIA 119 Proceedngs of the Internatonal Conference on Informaton Technologes (InfoTech-2007) 21 st 22 nd September 2007, Bulgara vol. 2 INVESTIGATION OF VEHICULAR USERS FAIRNESS

More information

Chapter 2: Displaying and Summarizing Data. Excel Insert Tab Charts Group. Column and Bar Charts 8/22/2011

Chapter 2: Displaying and Summarizing Data. Excel Insert Tab Charts Group. Column and Bar Charts 8/22/2011 Chapter 2: Dsplayng and Summarzng Data Part 1: Dsplayng Data Wth Charts and Graphs Copyrght 2010 Pearson Educaton, Inc. Publshng as Prentce Hall 2-1 Excel Insert Tab Charts Group Copyrght 2010 Pearson

More information

HYPOTHESIS TESTING OF PARAMETERS FOR ORDINARY LINEAR CIRCULAR REGRESSION

HYPOTHESIS TESTING OF PARAMETERS FOR ORDINARY LINEAR CIRCULAR REGRESSION HYPOTHESIS TESTING OF PARAMETERS FOR ORDINARY LINEAR CIRCULAR REGRESSION Abdul Ghapor Hussn Centre for Foundaton Studes n Scence Unversty of Malaya 563 KUALA LUMPUR E-mal: ghapor@umedumy Abstract Ths paper

More information

EFFECT OF SEEPAGE FORCES ON NILE RIVER BANK STABILITY

EFFECT OF SEEPAGE FORCES ON NILE RIVER BANK STABILITY Seventh Internatonal Water Technology Conference Egypt 28-30 March 2003 EFFECT OF SEEPAGE FORCES ON NILE RIVER BANK STABILITY Medhat Azz Assoc. Prof., Secretary General, Nle Research Insttute, NWRC, Caro,

More information

Questions that we may have about the variables

Questions that we may have about the variables Antono Olmos, 01 Multple Regresson Problem: we want to determne the effect of Desre for control, Famly support, Number of frends, and Score on the BDI test on Perceved Support of Latno women. Dependent

More information

Dr. Maddah ENMG 400 Engineering Economy 08/05/09. Chapter 18 Sensitivity Analysis and Expected Value Decisions

Dr. Maddah ENMG 400 Engineering Economy 08/05/09. Chapter 18 Sensitivity Analysis and Expected Value Decisions Dr. Maddah ENMG 400 Engneerng Economy 08/05/09 Chapter 18 Senstvty Analyss and Expected Value Decsons Senstvty analyss Engneerng economy estmates of parameters such as costs and other cash flow are only

More information

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence 1 st Internatonal Symposum on Imprecse Probabltes and Ther Applcatons, Ghent, Belgum, 29 June 2 July 1999 How Sets of Coherent Probabltes May Serve as Models for Degrees of Incoherence Mar J. Schervsh

More information

THE TITANIC SHIPWRECK: WHO WAS

THE TITANIC SHIPWRECK: WHO WAS THE TITANIC SHIPWRECK: WHO WAS MOST LIKELY TO SURVIVE? A STATISTICAL ANALYSIS Ths paper examnes the probablty of survvng the Ttanc shpwreck usng lmted dependent varable regresson analyss. Ths appled analyss

More information

+ + + - - This circuit than can be reduced to a planar circuit

+ + + - - This circuit than can be reduced to a planar circuit MeshCurrent Method The meshcurrent s analog of the nodeoltage method. We sole for a new set of arables, mesh currents, that automatcally satsfy KCLs. As such, meshcurrent method reduces crcut soluton to

More information

Moment of a force about a point and about an axis

Moment of a force about a point and about an axis 3. STATICS O RIGID BODIES In the precedng chapter t was assumed that each of the bodes consdered could be treated as a sngle partcle. Such a vew, however, s not always possble, and a body, n general, should

More information

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL

More information

8.5 UNITARY AND HERMITIAN MATRICES

8.5 UNITARY AND HERMITIAN MATRICES CHAPTER 8 COMPLEX VECTOR SPACES 8. UNITARY AND HERMITIAN MATRICES Problems nvolvng dagonalzaton of complex matrces and the assocated egenvalue problems requre the concept of untary and Hermtan matrces.

More information

GENETIC ALGORITHM FOR PROJECT SCHEDULING AND RESOURCE ALLOCATION UNDER UNCERTAINTY

GENETIC ALGORITHM FOR PROJECT SCHEDULING AND RESOURCE ALLOCATION UNDER UNCERTAINTY Int. J. Mech. Eng. & Rob. Res. 03 Fady Safwat et al., 03 Research Paper ISS 78 049 www.jmerr.com Vol., o. 3, July 03 03 IJMERR. All Rghts Reserved GEETIC ALGORITHM FOR PROJECT SCHEDULIG AD RESOURCE ALLOCATIO

More information

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008 Rsk-based Fatgue Estmate of Deep Water Rsers -- Course Project for EM388F: Fracture Mechancs, Sprng 2008 Chen Sh Department of Cvl, Archtectural, and Envronmental Engneerng The Unversty of Texas at Austn

More information

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts Power-of-wo Polces for Sngle- Warehouse Mult-Retaler Inventory Systems wth Order Frequency Dscounts José A. Ventura Pennsylvana State Unversty (USA) Yale. Herer echnon Israel Insttute of echnology (Israel)

More information

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications

Descriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications CMSC828G Prncples of Data Mnng Lecture #9 Today s Readng: HMS, chapter 9 Today s Lecture: Descrptve Modelng Clusterng Algorthms Descrptve Models model presents the man features of the data, a global summary

More information

Inter-Ing 2007. INTERDISCIPLINARITY IN ENGINEERING SCIENTIFIC INTERNATIONAL CONFERENCE, TG. MUREŞ ROMÂNIA, 15-16 November 2007.

Inter-Ing 2007. INTERDISCIPLINARITY IN ENGINEERING SCIENTIFIC INTERNATIONAL CONFERENCE, TG. MUREŞ ROMÂNIA, 15-16 November 2007. Inter-Ing 2007 INTERDISCIPLINARITY IN ENGINEERING SCIENTIFIC INTERNATIONAL CONFERENCE, TG. MUREŞ ROMÂNIA, 15-16 November 2007. UNCERTAINTY REGION SIMULATION FOR A SERIAL ROBOT STRUCTURE MARIUS SEBASTIAN

More information

The Analysis of Outliers in Statistical Data

The Analysis of Outliers in Statistical Data THALES Project No. xxxx The Analyss of Outlers n Statstcal Data Research Team Chrysses Caron, Assocate Professor (P.I.) Vaslk Karot, Doctoral canddate Polychrons Economou, Chrstna Perrakou, Postgraduate

More information

Contingent Claims and the Arbitrage Theorem

Contingent Claims and the Arbitrage Theorem Contngent Clams and the Arbtrage Theorem Paul J. Atzberger Paul J. Atzberger Please send any comments to: atzberg@math.ucsb.edu Introducton No arbtrage prncples play a central role n models of fnance and

More information

v a 1 b 1 i, a 2 b 2 i,..., a n b n i.

v a 1 b 1 i, a 2 b 2 i,..., a n b n i. SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 455 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces we have studed thus far n the text are real vector spaces snce the scalars are

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

A multi-objective geometric programming model for optimal production and marketing planning. S. J. Sadjadi*

A multi-objective geometric programming model for optimal production and marketing planning. S. J. Sadjadi* Journal of Industral Engneerng Internatonal July 008, Vol. 4, No. 7, 33-37 Islamc Azad Unversty, South Tehran Branch A mult-objectve geometrc programmng model for optmal producton and marketng plannng

More information

Fisher Markets and Convex Programs

Fisher Markets and Convex Programs Fsher Markets and Convex Programs Nkhl R. Devanur 1 Introducton Convex programmng dualty s usually stated n ts most general form, wth convex objectve functons and convex constrants. (The book by Boyd and

More information

A linear recurrence sequence of composite numbers

A linear recurrence sequence of composite numbers LMS J Comput Math 15 (2012) 360 373 C 2012 Author do:101112/s1461157012001143 A lnear recurrence sequence of composte numbers Jonas Šurys Abstract We prove that for each postve nteger k n the range 2 k

More information

MATH 697 ITERATIVE PROPORTIONAL SCALING

MATH 697 ITERATIVE PROPORTIONAL SCALING MATH 697 ITERATIVE PROPORTIONAL SCALING REN BETTENDORF 1. Abstract The purpose of ths project s to show observatons taken from the Iteratve Proportonal Scalng algorthm when used wth data taken from Herarchcal

More information

Time Domain simulation of PD Propagation in XLPE Cables Considering Frequency Dependent Parameters

Time Domain simulation of PD Propagation in XLPE Cables Considering Frequency Dependent Parameters Internatonal Journal of Smart Grd and Clean Energy Tme Doman smulaton of PD Propagaton n XLPE Cables Consderng Frequency Dependent Parameters We Zhang a, Jan He b, Ln Tan b, Xuejun Lv b, Hong-Je L a *

More information

Risk Model of Long-Term Production Scheduling in Open Pit Gold Mining

Risk Model of Long-Term Production Scheduling in Open Pit Gold Mining Rsk Model of Long-Term Producton Schedulng n Open Pt Gold Mnng R Halatchev 1 and P Lever 2 ABSTRACT Open pt gold mnng s an mportant sector of the Australan mnng ndustry. It uses large amounts of nvestments,

More information

Testing and Debugging Resource Allocation for Fault Detection and Removal Process

Testing and Debugging Resource Allocation for Fault Detection and Removal Process Internatonal Journal of New Computer Archtectures and ther Applcatons (IJNCAA) 4(4): 93-00 The Socety of Dgtal Informaton and Wreless Communcatons, 04 (ISSN: 0-9085) Testng and Debuggng Resource Allocaton

More information

NON LINEAR PROGRAMMING Prof. Stephen Graves

NON LINEAR PROGRAMMING Prof. Stephen Graves NON LINEAR PROGRAMMING Prof. Stephen Graves In a lnear program, the constrants are lnear n the decson varables, and so s the objectve functon. In a non lnear program, the constrants and/or the objectve

More information

The Analysis of Covariance. ERSH 8310 Keppel and Wickens Chapter 15

The Analysis of Covariance. ERSH 8310 Keppel and Wickens Chapter 15 The Analyss of Covarance ERSH 830 Keppel and Wckens Chapter 5 Today s Class Intal Consderatons Covarance and Lnear Regresson The Lnear Regresson Equaton TheAnalyss of Covarance Assumptons Underlyng the

More information

Traffic-light a stress test for life insurance provisions

Traffic-light a stress test for life insurance provisions MEMORANDUM Date 006-09-7 Authors Bengt von Bahr, Göran Ronge Traffc-lght a stress test for lfe nsurance provsons Fnansnspetonen P.O. Box 6750 SE-113 85 Stocholm [Sveavägen 167] Tel +46 8 787 80 00 Fax

More information

GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM

GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM BARRIOT Jean-Perre, SARRAILH Mchel BGI/CNES 18.av.E.Beln 31401 TOULOUSE Cedex 4 (France) Emal: jean-perre.barrot@cnes.fr 1/Introducton The

More information

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services When Network Effect Meets Congeston Effect: Leveragng Socal Servces for Wreless Servces aowen Gong School of Electrcal, Computer and Energy Engeerng Arzona State Unversty Tempe, AZ 8587, USA xgong9@asuedu

More information

Joint Network-Centric and User-Centric Radio Resource Management in a Multicell System*

Joint Network-Centric and User-Centric Radio Resource Management in a Multicell System* Jont Networ-Centrc and User-Centrc Rado Resource Management n a Multcell System* Nan Feng, Sun-Chuon Mau and Narayan B. Mandayam WINLAB, Dept. of ECE, Rutgers Unversty, Pscataway, NJ 08854 e-mal: {vonnan,

More information

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT Chapter 4 ECOOMIC DISATCH AD UIT COMMITMET ITRODUCTIO A power system has several power plants. Each power plant has several generatng unts. At any pont of tme, the total load n the system s met by the

More information

U.C. Berkeley CS270: Algorithms Lecture 4 Professor Vazirani and Professor Rao Jan 27,2011 Lecturer: Umesh Vazirani Last revised February 10, 2012

U.C. Berkeley CS270: Algorithms Lecture 4 Professor Vazirani and Professor Rao Jan 27,2011 Lecturer: Umesh Vazirani Last revised February 10, 2012 U.C. Berkeley CS270: Algorthms Lecture 4 Professor Vazran and Professor Rao Jan 27,2011 Lecturer: Umesh Vazran Last revsed February 10, 2012 Lecture 4 1 The multplcatve weghts update method The multplcatve

More information

Enabling P2P One-view Multi-party Video Conferencing

Enabling P2P One-view Multi-party Video Conferencing Enablng P2P One-vew Mult-party Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract Mult-Party Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P

More information

Lecture Lectur 12 Oct 24th 2008

Lecture Lectur 12 Oct 24th 2008 Lecture 12 Oct 24 th 2008 Revew Lnear SVM seeks to fnd a lnear decson boundary that maxmzes the geometrc t margn Usng the concept of soft margn, we can acheve tradeoff between maxmzng the margn and fathfully

More information

Statistical Methods to Develop Rating Models

Statistical Methods to Develop Rating Models Statstcal Methods to Develop Ratng Models [Evelyn Hayden and Danel Porath, Österrechsche Natonalbank and Unversty of Appled Scences at Manz] Source: The Basel II Rsk Parameters Estmaton, Valdaton, and

More information

Topic 8. Analysis of Variance (Ch. 12)

Topic 8. Analysis of Variance (Ch. 12) Topc 8. Analyss of Varance (Ch. 1) 1) One-way analyss of varance (ANOVA) 1.1 General ratonale We studed the comparson of two means n the last chapter, and found that we could use the t and/or z statstcs

More information

One-Way Analysis of Variance (ANOVA)

One-Way Analysis of Variance (ANOVA) One-ay Analyss of Varance (ANOVA) e are examnng dfferences between populaton means 1,,, k of k dfferent populatons where k s usually bgger than. Our goal s to determne whether there s any dfference between

More information

Estimation of Gini coefficients using Lorenz curves

Estimation of Gini coefficients using Lorenz curves Journal of Statstcal and Econometrc Methods, vol.1, no.2, 2012, 31-38 ISSN: 2241-0384 (prnt), 2241-0376 (onlne) Scenpress Ltd, 2012 Estmaton of Gn coeffcents usng Lorenz curves Johan Fellman 1,2 Abstract

More information

Damage detection in composite laminates using coin-tap method

Damage detection in composite laminates using coin-tap method Damage detecton n composte lamnates usng con-tap method S.J. Km Korea Aerospace Research Insttute, 45 Eoeun-Dong, Youseong-Gu, 35-333 Daejeon, Republc of Korea yaeln@kar.re.kr 45 The con-tap test has the

More information

Chapter 4 Least-Mean-Square Algorithm ( LMS Algorithm )

Chapter 4 Least-Mean-Square Algorithm ( LMS Algorithm ) Chapter 4 Least-Mean-Square Algorthm ( LMS Algorthm ) 1. Search Methods he optmum tap-weghts of a transversal (FIR) Wener flter can be obtaned by solvng the Wener-Hopf equaton provded that the requred

More information

SOLVING CARDINALITY CONSTRAINED PORTFOLIO OPTIMIZATION PROBLEM BY BINARY PARTICLE SWARM OPTIMIZATION ALGORITHM

SOLVING CARDINALITY CONSTRAINED PORTFOLIO OPTIMIZATION PROBLEM BY BINARY PARTICLE SWARM OPTIMIZATION ALGORITHM SOLVIG CARDIALITY COSTRAIED PORTFOLIO OPTIMIZATIO PROBLEM BY BIARY PARTICLE SWARM OPTIMIZATIO ALGORITHM Aleš Kresta Klíčová slova: optmalzace portfola, bnární algortmus rojení částc Key words: portfolo

More information

The covariance is the two variable analog to the variance. The formula for the covariance between two variables is

The covariance is the two variable analog to the variance. The formula for the covariance between two variables is Regresson Lectures So far we have talked only about statstcs that descrbe one varable. What we are gong to be dscussng for much of the remander of the course s relatonshps between two or more varables.

More information

Credit Limit Optimization (CLO) for Credit Cards

Credit Limit Optimization (CLO) for Credit Cards Credt Lmt Optmzaton (CLO) for Credt Cards Vay S. Desa CSCC IX, Ednburgh September 8, 2005 Copyrght 2003, SAS Insttute Inc. All rghts reserved. SAS Propretary Agenda Background Tradtonal approaches to credt

More information

MAPP. MERIS level 3 cloud and water vapour products. Issue: 1. Revision: 0. Date: 9.12.1998. Function Name Organisation Signature Date

MAPP. MERIS level 3 cloud and water vapour products. Issue: 1. Revision: 0. Date: 9.12.1998. Function Name Organisation Signature Date Ttel: Project: Doc. No.: MERIS level 3 cloud and water vapour products MAPP MAPP-ATBD-ClWVL3 Issue: 1 Revson: 0 Date: 9.12.1998 Functon Name Organsaton Sgnature Date Author: Bennartz FUB Preusker FUB Schüller

More information

Software project management with GAs

Software project management with GAs Informaton Scences 177 (27) 238 241 www.elsever.com/locate/ns Software project management wth GAs Enrque Alba *, J. Francsco Chcano Unversty of Málaga, Grupo GISUM, Departamento de Lenguajes y Cencas de

More information

Realistic Image Synthesis

Realistic Image Synthesis Realstc Image Synthess - Combned Samplng and Path Tracng - Phlpp Slusallek Karol Myszkowsk Vncent Pegoraro Overvew: Today Combned Samplng (Multple Importance Samplng) Renderng and Measurng Equaton Random

More information

Lecture 3.4 Electric Potential

Lecture 3.4 Electric Potential Lecture 3.4 Electrc Potental Today we are gong to look at electrostatc problems from a dfferent stand pont. We wll use the same dea whch we have developed n classcal mechancs. As you may recall, we frst

More information

Lecture 3. 1 Largest singular value The Behavior of Algorithms in Practice 2/14/2

Lecture 3. 1 Largest singular value The Behavior of Algorithms in Practice 2/14/2 18.409 The Behavor of Algorthms n Practce 2/14/2 Lecturer: Dan Spelman Lecture 3 Scrbe: Arvnd Sankar 1 Largest sngular value In order to bound the condton number, we need an upper bound on the largest

More information