. Introducton. The system of unsteady compressble Naver-Stokes (N.-S.) equatons s a fundamental system n ud dynamcs. To be able to solve the system qu

Size: px
Start display at page:

Download ". Introducton. The system of unsteady compressble Naver-Stokes (N.-S.) equatons s a fundamental system n ud dynamcs. To be able to solve the system qu"

Transcription

1 VARIABLE DEGREE SCHWARZ METHODS FOR THE IMPLICIT SOLUTION OF UNSTEADY COMPRESSIBLE NAVIER-STOKES EQUATIONS ON TWO-DIMENSIONAL UNSTRUCTURED MESHES Xao-Chuan Ca y Department of Computer Scence Unversty of Colorado Boulder, CO 839 ca@cs.colorado.edu Charbel Farhat Department of Aerospace Engneerng Unversty of Colorado Boulder, CO 839 charbel@alexandra.colorado.edu Marcus Sarks z Department of Computer Scence Unversty of Colorado Boulder, CO 839 msarks@cs.colorado.edu Abstract We report our experences on usng a new varant of the Schwarz precondtoned GM- RES methods n the mplct soluton of the unsteady compressble Naver-Stokes equatons dscretzed on two-dmensonal unstructured meshes. We rst partton the global mesh wth the recursve spectral bsecton method nto submeshes, and then we ntroduce a famly of Schwarz methods, referred to as the Varable Degree Schwarz methods (VDS) on the overlappng submeshes. In VDS, the subdoman precondtoner s constructed by usng a polynomal n two matrx varables, namely the matrx, n ts un-factorzed form, of the current tme step k and another matrx, n ts factorzed form, obtaned at a prevous tme step j. The degree of the matrx polynomal n each subdoman s determned automatcally so that extra precondtonng s performed only n subdomans whose assocated local matrces have large condton numbers. The extra precondtonng occurs often near the body of the arfol. We show numercally that VDS s very eectve. Unlke the well-known ellptc theory, we observe that the convergence rate of VDS precondtoned GMRES degenerates very mldly wthout a coarse space for reasonably large number of subdomans. We also study the eects of the overlappng sze, the number of subdomans and the level of nexactness of the subdoman solvers. The other purpose of the study s to understand the robustness of the Schwarz methods wth respect to ow parameters, such as the CFL, the free stream Mach number and the Reynolds number. Numercal results for both subsonc and transonc problems are reported. The work was supported n part by the NSF Grand Challenges Applcatons Group grant ASC and the NASA HPCC Group grant NAG y The work was supported n part by the NSF grant ASC , and by NASA Contract No. NAS- 948 whle the author was n resdence at the Insttute for Computer Applcatons n Scence and Engneerng (ICASE), NASA Langley Research Center, Hampton, VA z The work was supported n part by the NSF grant ASC

2 . Introducton. The system of unsteady compressble Naver-Stokes (N.-S.) equatons s a fundamental system n ud dynamcs. To be able to solve the system quckly and accurately n complex geometry s one of the ultmate goals of computatonal ud dynamcs [2]. To acheve the goal, several mportant technques have been developed n the past few years, such as unstructured grd generatons for complex geometry, stable, conservatve dscretzatons of the N.-S. equatons, unstructured grd parttonngs, as well as the powerful, robust mplct precondtoned teratve solvers dscussed below. Most of the technques are developed for steady state calculatons, see e.g., [, 5, 6] and references theren. In ths paper, we put all the technques together and ntroduce a robust doman decomposton based fast precondtoned teratve method for the tme accurate soluton of unsteady problems. We study mplct methods for solvng unsteady N.-S. equatons dscretzed on twodmensonal unstructured meshes wth a combned nte element/nte volume scheme for the spatal varables and a smple backward Euler scheme for the temporal varable. It s well-known that the man advantage of mplct methods s that they allow the tme steps to be determned solely based on the physcs of the ud ow, not on the stablty property of the tme dscretzaton scheme, [4, 3, 3]. However, to advance n tme, a large, sparse, nonsymmetrc lnear system of equatons must be constructed and solved at each tme step. Dependng of the sze of the tme step, and several other ow parameters, the condtonng of the matrx may change from well-condtoned to mldly ll-condtoned. And due to the complexty of the ow pattern, at a gven tme step, the matrx may be llcondtoned n certan subregons near the arfol and relatvely well-condtoned elsewhere. Detals about the local condtonng of the matrx wll be dscussed later. To solve these systems teratvely, t s necessary to have a famly of precondtoners whose strength can be adjusted locally n each subdoman accordng to the ow condton. Overlappng Schwarz methods (OSM) s a famly of precondtoners for solvng large sparse lnear systems arsng from the dscretzaton of partal derental equatons, see e.g. [6, 8,,27]. It was orgnally desgned for scalar lnear ellptc problems. We nd OSM to be very attractve for our purpose because () they are more parallelzable than the popularly used global ILU precondtoners; (2) they are ecent for nonsymmetrc and ndente problems; (3) they have mesh ndependent convergence rates, at least for ellptc nte element problems; (4) they have adjustable strength controlled by usng the nexact soluton technques for solvng local problems. We shall further explore the exblty of OSM at the subdoman level and ntroduce a new varant below. It s well-known that when constructng a precondtoner for solvng a sngle system of lnear equatons, Au = f, all the nformaton needs to be from the matrx A. However, the ssue for tme dependent problems s derent. A sequence of nter-related systems A (k) u = f (k) has to be solved. If the matrx, and especally n ts (often nexactly) factorzed form, obtaned at a prevous tme step can be properly used, then the precondtoner at the current tme step can be obtaned cheaply. More precsely speakng, at each tme step, we solve the global lnear system by a precondtoned GMRES method ([26]) and n the precondtonng stage, followng the general overlappng Schwarz framework, we solve the local subdoman problems by another precondtoned GMRES method, wth derent precondtoners and stoppng condtons. In each subdoman the precondtoner s bult by usng a polynomal n two matrx varables, namely the matrx, n ts un-factorzed form, of the current tme step k and another matrx, n ts factorzed form, obtaned at a prevous tme step j. The degree of the matrx polynomal reects the condtonng of the subdoman matrx. Note that classcal Schwarz methods correspond to the case where the degree of

3 the matrx polynomals always equals to one. In our method, the degree of the polynomal vares from subdoman to subdoman dependng the ow condtons, and therefore we refer to the methods as varable degree Schwarz methods (VDS). In ths paper, we also nvestgate the derence between the Schwarz famly of precondtoners and other methods such as the smple Jacob teratve method and the global ILU precondtoned teratve method. Wthn the Schwarz precondtoners, we try to understand the role of the overlappng sze between subdomans, the eect of the number of subdomans, and the eect of the nexact subdoman solvers. Snce the constructon of the precondtoner s expensve, we explore the possblty of re-usng the precondtoner for several tme steps. We restrct our attenton to sequental computers, and sngle level Schwarz algorthms. For steady state problems, some studes can be found n [8]. For other recent developments n unsteady calculatons, we refer the readers to [2, 3, 3]. Our focus s on the Schwarz algorthms, therefore only the smplest tme dscretzaton, namely the rst order backward Euler scheme, s consdered n ths paper. Hgher order schemes and ther nuence on the Schwarz precondtoners wll be dscussed n a forthcomng paper. The physcal model we choose to test our algorthms contans a sngle element NACA2 arfol at a rather large angle of attack wth a modest Reynolds number. Both subsonc and transonc cases are studed n the paper. The paper s organzed as follows. In x2, we dscuss the unsteady compressble N.-S. equatons n the conservatve form, the boundary condtons and a dscretzaton scheme. In x3, we study a precondtoned teratve method and ntroduce a new varant of the overlappng Schwarz precondtoners. Numercal experments for both subsonc and transonc ows are reported n x4. x5 ncludes a few nal remarks. 2. The two-dmensonal unsteady N.-S. equatons. In ths secton, we descrbe the two-dmensonal unsteady compressble N.-S. equatons n ts conservatve form. We also dscuss the spatal and temporal dscretzatons of the equatons on unstructured meshes. Followng the notons of Farhat, Fezou and Lanter [2, 3], and Fezou and Stouet [5], for the spatal varables, we use a combned nte element/nte volume scheme and for the temporal varable we use a smple backward Euler method. The scheme s of second order n space and rst order n tme. 2.. Governng equatons. Let < 2 be the ow doman and ts boundary, as shown n Fg.. The conservatve form of the N.-S. equatons s gven + r ~x F(W(~x;t)) = Re r ~x R(W(~x; t)); where ~x and t denote the spatal and temporal varables, and W = u v E C A ; r ~x @y C A ; F(W)= F (W) F 2 (W) Here the functons F and F 2 denote the convectve uxes F (W )= u u 2 + p uv v(e + p) C A ; F 2(W )= 2! ; R(W)= R (W) R 2 (W) v uv v 2 + p v(e + p) C A! :

4 and the functons R and R 2 denote the dusve uxes R (W )= xx xy u xx + v xy + k " Pr x C A ; R 2(W) = xy yy xy + v yy + k " Pr y In the above expressons, s the densty, U ~ =(u; v) T s the velocty vector, E s the total energy per unt of volume, and p s the pressure. These varables are related by the state equaton for perfect gas p =( ) E 2 k~ Uk 2 ; where denotes the rato of specc heats ( =:4, for ar). The specc nternal energy " s related to the temperature va " = c v T = E 2 k~ Uk 2 : In the dusve uxes, xx ; xy, and yy are the components of the two-dmensonal Cauchy stress tensor, k s the normalzed thermal conductvty, Pr = c p =k s the Prandtl number (Pr =:72, for ar), and Re = U ~ L = s the Reynolds number, where, U ~, L, and denote the characterstc densty, velocty, length, and dusvty, respectvely. The components of Cauchy stress tensor are related to the velocty va xx = 2 ; yy = 2 where denotes the C A ; xy Boundary condtons. We are nterested n unsteady, external ows around an arfol as pctured n Fg.. The doman boundary s = w [ and the far eld velocty s U ~. On the wall boundary w, a no-slp condton on U ~ and a Drchlet condton on the temperature T are mposed,.e., (2) ~U = ~; and T = T w : No boundary condton s speced for the densty. In the far eld, the vscous eect s assumed to be neglgble, therefore a unform free-stream velocty U ~ s mposed on (3) =; ~ U = cos sn! ; and p = ; M 2 where s the angle of attack, and M s the free-stream Mach number Dscretzaton. Let the temporal varable t be dscretzed as t k = t k + t k, where t k s the dscrete tme ncrement and t =. We consder the ncrement W k = W k W k, where W k s an approxmaton of W (;t k ). Note that when an algorthm s wrtten n the \delta" form [3, 28], the ncrement W k s the unknown varable rather than W k. Here, we use a rst-order nte derence approxmaton for the temporal varable, namely, the backward Euler scheme gven as (4) W k t k +(r W(F k )r ~x )W k Re (r W (R k 3 ) r ~x )W k

5 w ~U Fg.. The computatonal doman and ts wall and far eld boundares. = r( F k )+ Re rrk ; where F k and R k are approxmatons of F(W (;t k )) and R(W (;t k )), respectvely. We determne the tme step sze n the followng way. Let CFL be a pre-selected postve number. For each element, wth sze h, of the nte element mesh, we dene an element tme step sze by and then the global tme step s dened by t k = h CFL (C + ku k 2 )+2=(Re P r h ) (5) t k = mnft k g: Here C s the element sound speed, and U s the element velocty vector. The computatonal doman s dscretzed by a trangular grd as pctured n Fg.2. We use unstructured grds snce they provde exblty for tessellatng about complex geometres and for adaptng to ow features, such as shocks and boundary layers. We locate the varables at the vertces of the grd, whch gves rse to a cell-vertex scheme. The space of solutons s taken to be the space of pecewse lnear contnuous functons. The dscrete system s obtaned va a mxed Galerkn nte element/nte volume formulaton; see Farhat et al. [2, 3], and Fezou and Stouet [5] for detals. In short, the dscretzed system for (4) s obtaned as follows: For the tme dervatve of (4) we use a \mass-lumpng" technque, n whch we replace the mass matrx by some dagonal matrx. For the convectve terms of the left-hand sde of (4), we use a rst order scheme that s an extenson of Van Leer's MUSCL [29] scheme to the case of unstructured grds (see Fezou and Stouet [5]) wth a Roe approxmate Remann solver [24]. For the dusve terms of the left-hand sde of (4), we use a Galerkn nte element (rst-order quadrature ntegraton). For the convectve terms of the rght-hand sde of (4), we use a second-order scheme that s agan an extenson of Van Leer's MUSCL scheme to the case of unstructured grds (see Fezou and Stouet [5]) wth a Roe approxmate Remann solver [24]. We also use Van Albada's lmtng procedure to reduce numercal oscllatons of the solutons. For the dusve terms of the rght-hand sde of (4), we use a regular Galerkn nte element method (second-order quadrature ntegraton). 4

6 Fg. 2. The top left gure shows the un-decomposed nonunform shape regular nte element mesh. The top rght gure shows the decomposton of the mesh nto non-overlappng subdomans, and the bottom gure s a blow-up of the gure around the arfol. 5

7 Although both approxmatons we use for the left-hand sde of (4) are spatally rstorder, they operate on the ncrement W k and as a consequence the resultng scheme s spatally second-order for any xed t. We assume W k satses strongly the wall boundary condtons on w and the far eld boundary condtons at nnty. For the ntal values, W satses strongly the wall boundary condtons on w and the far eld boundary condtons at. For the ntal values and at the nteror nodes of, W takes the free-stream boundary condton. Puttng peces of the dscretzaton together, at each tme step, we obtan a large, sparse, generally nonsymmetrc lnear system of equatons of the form (6)! D (k) + B(k) t k u k = f k ; where D (k) s a dagonal, lumped mass matrx, B (k) s the sum of the convectve and dusve terms on the left-hand sde of (4) dscretzed by the nte volume and nte element methods, respectvely, f k s the dscretzed rght-hand sde of (4), and u k s the approxmate nodal value of W k at the nte element mesh ponts. We solve (6) by a precondtoned Krylov teratve method wth a precondtoner M (k) to a certan lnear tolerance,.e., (7) " M (k)! D (k) + B(k) t k u k f k 2 km (k) f k k 2 : To smplfy the dscusson, we shall use A (k) = D(k) + t k B (k) n the rest of the paper. The precondtoner M (k) wll be ntroduced n the next secton. 3. Soluton methods and varable degree Schwarz precondtoners. In ths secton, we rst brey recall the classcal addtve and multplcatve Schwarz algorthms. Then we ntroduce a varable degree verson of Schwarz algorthms that s more sutable for solvng tme dependent problems. Unlke steady state problems, f certan nformaton or matrx factorzaton obtaned at the prevous tme steps can be used a great deal of calculatons can be saved. Let us revew the Schwarz algorthms. Suppose that, at tme step k, we need to solve a lnear system of equatons A (k) u k = f k ; where A (k) s an explctly constructed, nonsymmetrc and sparse matrx wth symmetrc non-zero pattern. Snce our man nterest s on the multcomponent N.-S. equatons n twodmensonal spaces, each element ofa (k) can be consdered as a small 44 matrx, and each unknown of the vector u (k) s a `4-sze' vector. Thus, there s a bjecton between unknowns and vertces. We denote the set of vertces (or nodes) by N = f; ;ng, where n represents the total number of nodes (or unknowns). To dene algebrac Schwarz algorthms, see e.g. [6], we rst partton the set N nto n nonoverlappng subsets fn g whose unon s N.We use the TOP/DOMDEC mesh parttonng package of Farhat et al. [4] to obtan sets N. We use the recursve spectral bsecton method ([22]) wth certan optmzaton to obtan roughly the same number of nteror and boundary nodes n each N, and also to obtan good aspect rato on the subgrds. To generate an overlappng parttonng wth overlap 6

8 ovlp, we further expand each subgrd N by ovlp number of neghborng nodes, denoted as ~N. We denote by L the vector space spanned by the set ~N. For each subspace L,we dene an orthogonal projecton operator I as follows: I s a n n block dagonal matrx whose elements are 4 4 dentty matrces f the correspondng nodes belong to ~N and to 4 4 zero matrces otherwse. Wth ths we dene A (k) = I A (k) I ; whch s an extenson to the whole subspace, of the restrcton of A (k) to L. Note that although A (k) s not nvertble n the full space, ts restrcton to the subspace spanned by ~N s nonsngular, and we dene ts nverse by (A (k) ) I (A (k) ) jl I : The classcal addtve and multplcatve Schwarz algorthms can now be smply descrbed as follows: Solve the lnear system by a Krylov subspace method, where MA (k) u = Mf (k) (8) M =(A (k) ) ++(A (k) n ) ; for the addtve Schwarz algorthm, and (9) MA (k) =I I (A (k) ) A (k) I (A (k) n ) A (k) for the multplcatve Schwarz algorthm. It has been shown that the above mentoned algorthms are very successful for steady state CFD problems, see e.g., [5, 2] and also [9]. There are three major steps n the constructon of the Schwarz precondtoners, namely ) the constructon of the matrx A (k) ; 2) the constructon of the matrces A (k) ; and 3) the ncomplete factorzaton of the matrces A (k). In fact Step ) s not necessary snce the matrces constructed n Step 2) can be used to calculate the matrx-vector multplcatons. Snce we are nterested n mplct methods, Step 2) has to be done at every tme step no matter how expensve t s. One expensve step n the constructon of the precondtons as formulated above for tme dependent problems One way toavod the frequent factorzaton of A (k) s to smply use some calculated at tme step j, where j<k. However, ths method may not be very eectve fjand k are too far apart. More dscussons on usng frozen precondtoners can be found later n the paper. Another problem wth the Schwarz precondtoners (8) and (9) s that all subdomans are treated equally n terms of the level of precondtonng n the sense that the number of s Step 3). old factorzed matrx A (j) applcatons of (A (k) ), or ts nexact verson, s the same on all subdomans, regardless of the fact that the subdoman matrces A (k) have vary derent condton numbers. Physcally speakng, the behavor of the partal derental equatons n subdomans near the body of the arfol, or near the shocks s very derent from the regons that are far from the subdomans where the real actons take place. 7

9 Here we propose a method that places derent level of precondtonng n derent subdomans and wll also show by numercal experments that the methods reman eectve even f j and k are far apart from each other. The dea s smple. We replace the requred matrx-vector multply n (8) or (9) () w =(A (k) ) v by another teratve procedure wth (B (j) ) as the precondtoner. Here B (j) s an ncomplete factorzaton of A (j) wth certan levels of ll-ns. More precsely speakng, to obtan w for a gven v, we run several steps of GMRES n the subspace L to drve the resdual () (B (j) ) (v A (k) ~w) 2 (B (j) We then set w := ~w. Here s pre-selected small value. Examples can be found n x4 of ths paper. In the matrx language, we replace the matrx (A (k) ) n () by a matrx polynomal poly (B (j) ) A (k) of a certan degree, whch depends of the number of GMRES teratons needed n the subspace L.To put them nto a sngle form, the addtve Schwarz precondtoner becomes ) v : 2 M = poly (B (j) ) A (k) + +poly n (B (j) n ) A (k) n : Note that ths precondtoner does not contan (A (k) ), but t contans certan spectral nformaton of (A (k) ). Ths makes t very eectve. In fact, M s a truncated seres representaton of (A (k) ) based on a splttng of A (k) nto the sum of B (j) and A (k) B (j). A dscusson on a related polynomal precondtonng method can be found n [7]. We note that n a gven subdoman, the number of GMRES teratons, or the degree of the polynomal, s determned by the condtonng of the local stness matrx. The multplcatve verson can be constructed n a smlar way. The extenson to the local multplcatve Schwarz method ([7]) s also straghtforward. We remark that snce the precondtoner changes n the GMRES loop due to the stoppng condton determned by, t s generally more approprate to use the so-called exble GMRES [25], whch s slghtly more expensve than the regular one. We do not use the exble GMRES snce the regular GMRES presents no problem for our test cases. The mplementaton of the methods s rather complcated because of the use of mult-layered Krylov teratons as a global, or outer, and also local solvers. PETSc makes our numercal tests possble. More detals regardng to the mplementaton wll be dscussed n the next secton. 4. Numercal results. The goal of ths secton s to demonstrate the usefulness of the famly of VDS precondtoners n the mplct soluton of compressble ow problems, and to compare the eectveness of the methods wth varous other methods, such as the pontwse Jacob teratve method and the global ILU() precondtoned GMRES method, for both subsonc and transonc ows. The experments were performed on a DEC Alpha workstaton (275MHz, 52MB memory), and the software was wrtten by usng the newly developed package PETSc [9] of the Argonne Natonal Laboratory. All arthmetc operatons are n 8

10 double precson. The system BLAS lbrary (dxml []) was used. Only sequental results are reported here. A parallel verson of the code s beng developed, and the results wll be reported n the future. Here we apply our computatonal algorthms to the smulaton of two-dmensonal low Reynolds number chaotc ows past a NACA2 arfol at hgh angle of attack and two derent Mach numbers. It was shown n Pullam [23] that such ows can be resolved wth a reasonable number of grd ponts. The accuracy of the computed solutons are substantated by successve mesh renements and comparsons wth the results were reported n [23]. No steady state solutons exsts for both test cases descrbed below. Test : The subsonc case wth free stream Mach number M =: and Re = 8:. We use a pre-generated shape regular trangular mesh wth 228 nodes; see Fg.2 for example. The Mach surfaces of the computed soluton at varous tme steps are gven n Fg.5. Test 2: The transonc case wth free stream Machnumber M =:84 and Re = 6:. We use a mesh wth nodes obtaned by unformly renng the mesh used n Test. The Mach surfaces of the computed soluton at varous tme steps are gven n Fg.6. Table Total CPU hours and tme steps spent for calculatng the lft curves usng explct and mplct methods wth derent CFL numbers. The total non-dmensonalzed tme nterval s (, 25) for the M =:case and (, ) for the M =:84 case. Expl. Impl. Impl. Impl. CFL=.8 CFL=25 CFL=5 CFL= M =: CPU(hours) Mesh2k Tme steps M =:84 CPU(hours) Mesh48k Tme steps In the followng dscussons, we shall refer to these two meshes as \Mesh2k" and \Mesh48k", respectvely. In the mplementaton of Schwarz precondtoners, we partton the mesh by usng the TOP/DOMDEC package [4], whch mplements the recursve spectral bsecton method. We requre that all subdomans have more or less the same number of mesh ponts. An eort s made to reduce the number of mesh ponts along the nterfaces of subdomans, whch may be needed later n our parallel code to reduce the communcaton cost. The mesh generaton and parttonng are consdered as pre-processng steps, and therefore not counted toward the CPU tme reported n Table. The sparse matrx dened by (4) s constructed at every tme step, and stored n the Compressed Sparse Row format. The subdoman matrces are obtaned by takng elements, accordng to a pre-selected ndex set, from the global matrx. A symbolc ILU() factorzaton of the subdoman matrx s performed at the very rst tme step, and re-used at all the later tme steps. Ths s possble due to the fact that the matrces, constructed at every tme step, share the same non-zero pattern. We also tested the ILU(k) (k>) precondtoners, whch are not compettve wth ILU() n terms of the CPU tme n our mplementaton for both test cases. We remark that f ILU wth drop tolerance s used then the non-zero pattern of the matrces may change and therefore the symbolc factorzaton may not be very useful. We note that at the begnnng of the ow movement,.e., when the non-dmensonalzed tme t :, the ow changes so drastcally that the use of any t n that makes the correspondng CFL number larger than : would result n the loss of tme accuracy for the 9

11 3 Mach., Re 8, Angle of Attack 3, Mesh2k 3 Mach.84, Re 6, Angle of Attack 3, Mesh48k + > Explcty scheme wth CFL.8 + > Explcty scheme wth CFL x > Implcty scheme wth CFL x > Implcty scheme wth CFL 25 o > Implcty scheme wth CFL 5 o > Implcty scheme wth CFL 5 2 * > Implcty scheme wth CFL 2 * > Implcty scheme wth CFL Lft.5 Lft Non Dmensonalzed Tme Non Dmensonalzed Tme Fg. 3. Lft curves obtaned by explct and mplct methods. The left gure shows the subsonc cases wth M =:,Re = 8: and the number of mesh ponts s 228. The rght gure s for the transonc cases wth M =:84, Re = 6: and the number of mesh ponts s entre calculaton. Ths mples that small t n have to be used when t :, and therefore, the mplct method has to be abandoned for ths ntal perod of tme. In our experments, the mplct solver s turned on at t =:. The soluton for the perod <t:s obtaned wth the explct method wth CFL= The eect of usng large CFL numbers. One of the bggest advantages of mplct methods s that large tme steps, or large CFL numbers, can be used. We rst examne ths clam by comparng the tme accuracy of two solutons obtaned by usng the mplct methods wth CFL=25, 5 and, respectvely, and the soluton obtaned by a second order (n tme and space) explct method, [2]. From Fg.3, one can easly see that no two derent CFL numbers gve dentcal solutons. However, the mportant thng for engneerng purposes s to capture correctly the \phase" and \ampltude" of the lfts. Small errors n ampltude and phases are usually admssble. Ths can often substantally shorten the turnaround tmes n the ntal aerodynamc desgn and analyss. For the mplct methods, we use the multplcatve VDS precondtoned GMRES method as the global lnear solver. The subdoman precondtoners are re-computed at every 5 tme steps. In the Schwarz methods, we use 8 subdomans, and ovlp =. Each subdoman problem s solved by an ILU() precondtoned GMRES method wth the stoppng tolerance set to =. The stoppng tolerance for the global lnear system solver s = 3. We also test several cases wth smaller, such as 4 and 5. The resultng lft curves are not dstngushable from the ones shown n Fg.3. In Table, we report the total number of tme steps and CPU tme n hours spent on the entre computaton, not ncludng the mesh generaton and parttonng. We observe from our experments that even wth a CFL number,, not much tme accuracy s lost for a certan perod of tme, see Fg.3. However, f the tme accuracy s requred for a longer perod of tme, we do recommend a smaller CFL number. We remark that our dscussons here on the eects on usng large CFL numbers s based on our rst order tme dscretzaton scheme. The results may change slghtly f hgher order schemes are used The Schwarz parameters. The number of subdomans and the sze of overlap are two mportant parameters for Schwarz methods. We here test the addtve and multplcatve VDS methods for both test cases. Instead of runnng the entre calculaton as

12 we dd n the prevous secton, we run the tests for only tme steps startng at t =:. In terms of the non-dmensonal tme, ths means tme ntervals (; :86) for Test and (; 2:28) for Test 2. In the rest of the paper, we shall use MaxIt to denote the maxmum number of global GMRES teratons and TotalIt the total numbers of global GMRES teratons wthn ths lnear system solves. To measure the approxmate cost of the methods wthout ncludng any machne dependent factors, we use EMatVec to denote the equvalent number of matrx-vector multplcatons, whch ncludes the actual stness matrx-vector multplcatons and the precondtonng-matrx-vector multplcatons. Snce derent subdomans may need derent number of matrx-vector multplcatons, we take the average over all subdomans and convert t nto a multple of the equvalent global matrx-vector multplcatons. Suppose that ~n s the sze of the global matrx. Note that ~n s 4 tmes the number of mesh ponts. Then, one global matrx-vector multplcaton requres roughly 28~n ops. Our prmary teratve solver GMRES has a complexty of I(I+ 2)~n+ I(cost of a precondtoned matrx-vector multplcaton). Here I s the number of teratons. For example, the pure GMRES cost, wthout countng the cost of the matrx-vector multplcatons, for 4 teratons s about 24~n, whch s a lttle less than the cost of dong global stness matrx-vector multplcaton. Let us rst dscuss the dependence of the convergence rate on the number of subdomans. We use 5 derent decompostons of, wth both Mesh2k and Mesh48k. The number of subdomans goes from 8 to 28. We run both Test and 2, wth ovlp equals to one ne mesh cell. In Table 2, we present the maxmum number of global GMRES teratons wthn one hundred tme steps and ts correspondng EMatVec. If multplcatve VDS s used even wthout the specal subdoman colorng or orderng, MaxIt s ndependent of the number of subdomans, for reasonably large number of subdomans, such as 28. An nterestng case s shown on the top left porton of Table 2, whch ndcates that f addtve VDS s used for the subsonc problem, the number of maxmum teratons does grow, though not very fast, as the number of subdomans becomes large. In ths case, we beleve that a coarse space may be useful to reduce the dependence on the number of subdomans. However, we have not mplemented the coarse grd solver yet. For transonc problems, our tests show that the use of a coarse level grd s not necessary wth both addtve and multplcatve VDS precondtoners. Whether overlap s useful or not s a rather subtle ssue. It depends on the global lnear stoppng parameter dened n (7) and the local lnear stoppng parameter dened n (). In Table 3, we report the case for = 6 and varyng. Accordng to the results n Table 3 and a large number of other tests we dd (not beng reported here), large overlaps can reduce the number of teratons and CPU tme only f the stoppng parameter s small. In our stuaton when = 3,we nd = oers the best CPU results, and therefore we do not need large overlaps. In the rest of the tests, we shall use ths set of and, wth overlap. We next exam the VDS methods. We focus on the case wth 8 subdomans, and use GMRES/ILU() as the local subdoman solvers. The parttonngs used for Mesh2k and Mesh48k are derent. The subdomans are numbered as n Fg.4. The results obtaned for one hundreds tme steps startng at t =: are summerzed n Table 4. We have also run the test for t equals to other values and the results are more or less the same. We use the same local stoppng condton, namely = for all subdomans and for both subsonc and transonc problems. It turns out the requred degrees of local precondtonng

13 Fg. 4. The left gure shows the parttonng of Mesh2k nto 8 subdomans and the rght ones shows that for Mesh48k. polynomals are qute derent. For the subsonc case, subdomans 6 and 8 need more teratons (4 and 6 respectvely) than other subdomans. The left pcture of Fg.4 shows that these two subdomans cover the top porton of the arfol. Only two teratons are needed for subdomans that are far away from the arfol, such as, 2 and 3. The number of teratons reects the condtonng of the subdoman matrx. For the transonc case, all subdomans need ether one or two more teratons. For both test problems, Table 2 and Table 4 show that both the number of global teratons and the number of local teratons are surprsngly small, whch ndcate that the lnear system of equatons (6) s n fact not too ll-condtoned. We beleve that ths s due to the use of relatvely small tme steps (5) A comparson wth the pontwse Jacob teratve method. For comparson purpose, we solve both test problems by usng the smplest teratve method, namely the pontwse Jacob method, whch s often referred to as the Jacob precondtoned Rchardson's method. Jacob method has a few very attractve features, such as beng easy to parallelze. Note that for multcomponent test problems, a pont corresponds to a 4 4 matrx. When usng the Jacob method, a dense 4 4 matrx needs to be nverted at every mesh pont. In our experments, at each tme step before solvng the lnear system, we compute the nverse of these 4 4 matrces explctly and save them for the Jacoban teratons. As before, we run the tests for one hundred tme steps and record the maxmum number of teratons as well as the total number of teratons, see Table 5. We observe that, n terms of teraton numbers for solvng lnear systems, the transonc problem s easer to handle than the subsonc problem, whch s more of an ellptc system. Comparng the TotalIt, whch equals to the total number of EMatVec, n Table 5, and the total EMatVec numbers n Table 2, we found that Jacob s consderably more expensve than the Schwarz precondtoned GMRES methods. We beleve that the requred number of teratons would grow much faster f ner meshes are used than that of the Schwarz precondtoned GMRES methods. 2

14 4.4. A comparson wth a global ILU() precondtoned GMRES method. In Table 5, we show the maxmum and total number of teratons when usng the global ILU() precondtoned GMRES method for both test cases. In terms of the number of EMatVec, the method outperforms, by a factor of 5% to 3%(Comparng Table 5 and the bottom part of Table 2), the multplcatve VDS precondtoned GMRES methods we dscussed n the paper. Unfortunately, ts parallelzaton on dstrbuted memory computers s not very easy The eect of frozen precondtoner. Fnally, we examned the eect of usng the same precondtoner, or part of the precondtoner, for several tme steps wthout dong the factorzaton at every tme step. In Table 6, we summerze the results for usng derent numbers of frozen steps, namely Froz = ; 5; :::. There s a range of optmal Froz one can choose from; smlar numbers of EMatVec were obtaned n our mplementaton for Froz rangng from 5 to 5. For the subsonc case, we can go a bt further, e.g., take Froz =. 5. Conclusons. We proposed and tested a famly of varable degree Schwarz (VDS) precondtoned GMRES methods for solvng lnear systems that arse from the dscretzaton of unsteady, compressble N.-S equatons on 2D unstructured meshes for both subsonc and transonc ows past a sngle element NACA2 arfol. We found that wth mplct methods, larger tme steps can be used and the overall smulaton tme can be reduced sgncantly comparng wth the explct method. In VDS, the level of precondtonng n each subdoman vares accordng to the local ow condton, therefore extra precondtonng s performed only when and where t s needed. For subsonc problems, we found that the condtonng of the subdoman matrces changes qute a bt from one ow regon to another, and extra local precondtonng n subdomans n whch the ow changes drastcally can sgncantly reduce the total number of global lnear teratons. Ths s somewhat less obvous for transonc ow, whch needs a nearly unformly small global and local number of teratons. When usng VDS, the best results are obtaned wth small overlap. For the multplcatve verson, the convergence rate depends very mldly on the number of subdomans (up to 28 subdomans has been tested), and for the addtve verson, a slght dependence s observed for the subsonc test problem and therefore a coarse space mght be useful. We also compared our methods wth the smple pont(means 44 block for our multcomponent problems) Jacob teratve method and the global ILU() precondtoned GMRES method. We found that Jacob s sgncantly slower than the proposed methods, especally for the subsonc case, and f sequental computers are the prmary computng platform, then the global ILU() precondtoned GMRES s a wnner over all methods we have tested. All of our tests were done on a sequental computer. Consderable eort s needed n order to obtan a well-balanced parallel mplementaton. We remark that our mesh parttonng s obtaned wthout the knowledge of the ow condton and our experences show that a soluton dependent mesh parttonng would oer a more computatonally balanced decomposton of the problems. 3

15 Fg. 5. The Mach dstrbuton at M =:, for non-dmensonalzed tme t =2;3; :::; 9. 4

16 Fg. 6. The Mach dstrbuton at M =:84, for non-dmensonalzed tme t =2;3; :::; 9. 5

17 REFERENCES [] T. J. Barth, T. Chan, and W. P. Tang, Implct parallel precondtonng technques for computatonal ud dynamcs, n Proceedngs of the Copper Mountan Conference on Iteratve Methods, T. Manteuel and S. McCormck, eds., Copper Mountan, Colorado, 996. [2] T. J. Barth and S. W. Lnton, An unstructured mesh Newton solver for compressble ud ow and ts parallel mplementaton, AIAA Paper 95-22, (Jan. 995). [3] R. M. Bean and R. F. Warmng, An mplct nte-derence algorthm for hyperbolc systems n conservaton laws, J. Comp. Phys., 22 (976), pp. 87{. [4] X.-C. Ca, C. Farhat, and M. Sarks, Schwarz methods for the unsteady compressble Naver-Stokes equatons on unstructured meshes, n Doman Decomposton Methods n Scences and Engneerng, R. Glownsk, J. Peraux, Z. Sh, and O. Wdlund, eds., England, 996, John Wley & Sons, Ltd. To appear. [5] X.-C. Ca, W. D. Gropp, D. E. Keyes, R. G. Melvn, and D. P. Young, Parallel Newton- Krylov-Schwarz algorthms for the transonc full potental equaton, SIAM J. Sc. Comput., (996). Submtted. [6] X.-C. Ca and Y. Saad, Overlappng doman decomposton algorthms for general sparse matrces, Numer. Ln. Alg. Applcs, 3 (996), pp. {8. [7] X.-C. Ca and M. Sarks, Local multplcatve Schwarz algorthms for convecton-duson equatons, Tech. Rep. ICASE Report No , ICASE, NASA Langley Research Center, 995. [8] X.-C. Ca and O. Wdlund, Multplcatve Schwarz algorthms for nonsymmetrc and ndente ellptc problems, SIAM J. Numer. Anal., 3 (993), pp. 936{952. [9] T. F. Chan and T. P. Mathew, Doman decomposton algorthms, Acta Numerca, (994). [] Dgtal Equpment Corporaton, Dgtal Extended Math Lbrary for DEC OSF/ AXP: Reference Manual, Dgtal Equpment Corporaton, 993. [] M. Dryja and O. B. Wdlund, Doman decomposton algorthms wth small overlap, SIAM J. Sc. Comp., 5 (994), pp. 64{62. [2] C. Farhat, L. Fezou, and S. Lanter, Two-dmensonal vscous ow computaton on the Connecton Machne: Unstructured meshes, upwnd schemes and parallel computaton, Comput. Methods Appl. Mech. Engrg., 2 (993), pp. 6{88. [3] C. Farhat and S. Lanter, Smulaton of compressble vscous ows on a varety of MPPs: Computatonal algorthms for unstructured dynamc meshes and performances results, Comput. Methods Appl. Mech. Engrg., 9 (994), pp. 35{6. [4] C. Farhat, S. Lanter, and H. Smon, TOP/DOMDEC: A software tool for mesh parttonng and parallel processng and applcatons to CSM and CFD computatons, Comput. Sys. Engrg., 6 (995), pp. 3{26. [5] L. Fezou and B. Stoufflet, A class of mplct upwnd schemes for Euler smulatons wth unstructured meshes, J. Comp. Phys., 84 (989), pp. 74{26. [6] P. Forsyth and H. Jang, Iteratve methods for full Newton Jacoban for compressble Naver-Stokes equatons, n Proceedngs of the Copper Mountan Conference on Iteratve Methods, T. Manteuel and S. McCormck, eds., Copper Mountan, Colorado, 996. [7] G. Golub and J. M. Ortega, Scentc Computng: An Introducton wth Parallel Computng, Academc Press, Inc., 993. [8] W. D. Gropp, D. E. Keyes, and J. S. Mounts, Implct doman decomposton algorthms for steady, compressble aerodynamcs, n Sxth Conference on Doman Decomposton Methods for Partal Derental Equatons, A. Quarteron, J. Peraux, Y. A. Kuznetsov, and O. B. Wdlund, eds., Provdence, RI, 994, AMS. [9] W. D. Gropp, B. F. Smth, and L. C. McInnes, PETSc 2. User's Manual, Tech. Rep. ANL-95/, Argonne Natonal Laboratory, 995. [2] C. Hrsch, Numercal Computaton of Internal and External Flows, Vol I, Wley, New York, 99. [2] D. Knoll, Newton-Krylov-Schwarz methods appled to the Tokamak edge plasma ud equatons, n Doman-Based Parallelsm and Problem Decomposton Methods n Computatonal Scence and Engneerng, D. Keyes, Y. Saad, and D. Truhlar, eds., Phladelpha, Pennsylvana, 994, SIAM. [22] A. Pothen, H. D. Smon, and K.-P. Lou, Parttonng sparse matrces wth egenvectors of graphs, SIAM J. Matrx Anal. Appl., (99), pp. 43{452. [23] T. H. Pullam, Low Reynolds number numercal solutons of chaotc ows, AIAA Paper 89-23, (989). 27th Aerospace Scences Meetng, Reno, Nevada. [24] P. R. Roe, Approxmate Remann solvers, parameters vectors and derences schemes, J. Comp. Phys., 43 (98), pp. 357{37. 6

18 [25] Y. Saad, A exble nner-outer precondtoned GMRES algorthm, SIAM J. Sc. Stat. Comput., 4 (993), pp. 46{469. [26] Y. Saad and M. H. Schultz, GMRES: A generalzed mnmum resdual algorthm for solvng nonsymmetrc lnear systems, SIAM J. Sc. Stat. Comput., 7 (986), pp. 856{869. [27] B. F. Smth, P. E. Bjrstad, and W. D. Gropp, Doman Decomposton: Parallel Multlevel Methods for Ellptc Partal Derental Equatons, Cambrdge Unversty Press, 996. [28] J. Steger and R. F. Warmng, Flux vector splttng for nvscd gas dynamc wth applcaton to nte-derence methods, J. Comp. Phys., 4 (98), pp. 263{293. [29] B. van Leer, Towards the ultmate conservaton derence schemes V: A second-order sequel to Goudonov's method, J. Comp. Phys., 32 (979), pp. 36{37. [3] V. Venkatakrshnan, A perspectve on unstructured grd ow solvers, Tech. Rep. ICASE Report No. 95-3, ICASE, NASA Langley Research Center, Feb [3] V. Venkatakrshnan and D. J. Mavrpls, Implct method for the computaton of unsteady ows on unstructured grds, Tech. Rep. ICASE Report No. 95-6, ICASE, NASA Langley Research Center, Aug

19 Table 2 Each test s for tme steps and at each tme step the ntal precondtonedresdual s reduced by a factor of = 3 by usng GMRES/VDS wth ovlp =. We use GMRES/ILU() as nexact local solvers to reduce the local precondtoned resdual by a factor of =. Here CFL=5. ASM Test Test 2 subdomans MaxIt TotalIt EMatVec MaxIt TotalIt EMatVec MSM Test Test Table 3 GMRES teraton numbers to reduce the precondtoned resdual of Test to = 6 usng GM- RES/(addtve VDS) wth 8 subdomans. We use GMRES/ILU() as nexact local solver wth derent local stoppng crtera. Here CFL=. ovlp = ovlp = ovlp = 2 ovlp = 3 teraton =5:e =3:e =:e =:e exact 5 9 Table 4 The maxmum (MaxIt) and total (TotalIt) local GMRES/ILU() teraton numbers. The global solver s GMRES/(multplcatve VDS). The parameters are = 3,=, ovlp = and the local solvers are ILU(). Here CFL= Test, MaxIt TotalIt Test 2, MaxIt TotalIt

20 Table 5 The number of equvalent EMatVec operatons needed for tme steps startng at t =:. = 3, =, CFL= Jacob Global ILU() Test, MaxIt 74 EMatVec Test 2, MaxIt 46 4 EMatVec Table 6 The number of EMatVec operatons needed for tme steps startng at t =:. In GMRES/(multplcatve VDS), = 3,=, CFL=5, number of subdomans s 8 and ovlp =. For the Froz=2 case, the numbers are taken for 2 tme steps dvded by 2. Froz= Test, EMatVec TotalIt Test 2, EMatVec TotalIt

a tree-dmensonal settng. In te presented approac, we construct te ne mes by renng an exstng coarse mes and updatng te nodes of te ne mes accordng to t

a tree-dmensonal settng. In te presented approac, we construct te ne mes by renng an exstng coarse mes and updatng te nodes of te ne mes accordng to t Parallel Two-Level Metods for Tree-Dmensonal Transonc Compressble Flow Smulatons on Unstructured Meses R. Atbayev a, X.-C. Ca a, and M. Parascvou b a Department of Computer Scence, Unversty of Colorado,

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

Preconditioning Parallel Sparse Iterative Solvers for Circuit Simulation

Preconditioning Parallel Sparse Iterative Solvers for Circuit Simulation Precondtonng Parallel Sparse Iteratve Solvers for Crcut Smulaton A. Basermann, U. Jaekel, and K. Hachya 1 Introducton One mportant mathematcal problem n smulaton of large electrcal crcuts s the soluton

More information

RECENT research on structured mesh flow solver for aerodynamic problems shows that for practical levels of

RECENT research on structured mesh flow solver for aerodynamic problems shows that for practical levels of A Hgh-Order Accurate Unstructured GMRES Algorthm for Invscd Compressble Flows A. ejat * and C. Ollver-Gooch Department of Mechancal Engneerng, The Unversty of Brtsh Columba, 054-650 Appled Scence Lane,

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

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

Very simple computational domains can be discretized using boundary-fitted structured meshes (also called grids)

Very simple computational domains can be discretized using boundary-fitted structured meshes (also called grids) Structured meshes Very smple computatonal domans can be dscretzed usng boundary-ftted structured meshes (also called grds) The grd lnes of a Cartesan mesh are parallel to one another Structured meshes

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

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

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

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

Wavefront Reconstructor

Wavefront Reconstructor A Dstrbuted Smplex B-Splne Based Wavefront Reconstructor Coen de Vsser and Mchel Verhaegen 14-12-201212 2012 Delft Unversty of Technology Contents Introducton Wavefront reconstructon usng Smplex B-Splnes

More information

Hermite Splines in Lie Groups as Products of Geodesics

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

More information

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

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

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

More information

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

User Authentication Based On Behavioral Mouse Dynamics Biometrics

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

More information

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

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements Module 3: Element Propertes Lecture : Lagrange and Serendpty Elements 5 In last lecture note, the nterpolaton functons are derved on the bass of assumed polynomal from Pascal s trangle for the fled varable.

More information

arxiv: v3 [cs.na] 18 Mar 2015

arxiv: v3 [cs.na] 18 Mar 2015 A Fast Block Low-Rank Dense Solver wth Applcatons to Fnte-Element Matrces AmrHossen Amnfar a,1,, Svaram Ambkasaran b,, Erc Darve c,1 a 496 Lomta Mall, Room 14, Stanford, CA, 9435 b Warren Weaver Hall,

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

Solitary and Traveling Wave Solutions to a Model. of Long Range Diffusion Involving Flux with. Stability Analysis

Solitary and Traveling Wave Solutions to a Model. of Long Range Diffusion Involving Flux with. Stability Analysis Internatonal Mathematcal Forum, Vol. 6,, no. 7, 8 Soltary and Travelng Wave Solutons to a Model of Long Range ffuson Involvng Flux wth Stablty Analyss Manar A. Al-Qudah Math epartment, Rabgh Faculty of

More information

A One-Sided Jacobi Algorithm for the Symmetric Eigenvalue Problem

A One-Sided Jacobi Algorithm for the Symmetric Eigenvalue Problem P-Q- A One-Sded Jacob Algorthm for the Symmetrc Egenvalue Problem B. B. Zhou, R. P. Brent E-mal: bng,rpb@cslab.anu.edu.au Computer Scences Laboratory The Australan Natonal Unversty Canberra, ACT 000, Australa

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

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 HIGH-ORDER SPECTRAL (FINITE) VOLUME METHOD FOR CONSERVATION LAWS ON UNSTRUCTURED GRIDS

A HIGH-ORDER SPECTRAL (FINITE) VOLUME METHOD FOR CONSERVATION LAWS ON UNSTRUCTURED GRIDS AIAA-00-058 A HIGH-ORDER SPECTRAL (FIITE) VOLUME METHOD FOR COSERVATIO LAWS O USTRUCTURED GRIDS Z.J. Wang Department of Mechancal Engneerng Mchgan State Unversty, East Lansng, MI 88 Yen Lu * MS T7B-, ASA

More information

An Optimal Algorithm for Prufer Codes *

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

More information

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

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

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

Kent State University CS 4/ Design and Analysis of Algorithms. Dept. of Math & Computer Science LECT-16. Dynamic Programming

Kent State University CS 4/ Design and Analysis of Algorithms. Dept. of Math & Computer Science LECT-16. Dynamic Programming CS 4/560 Desgn and Analyss of Algorthms Kent State Unversty Dept. of Math & Computer Scence LECT-6 Dynamc Programmng 2 Dynamc Programmng Dynamc Programmng, lke the dvde-and-conquer method, solves problems

More information

USING GRAPHING SKILLS

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

More information

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

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

High-Boost Mesh Filtering for 3-D Shape Enhancement

High-Boost Mesh Filtering for 3-D Shape Enhancement Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,

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

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

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

NORMALE. A modied structured central scheme for. 2D hyperbolic conservation laws. Theodoros KATSAOUNIS. Doron LEVY

NORMALE. A modied structured central scheme for. 2D hyperbolic conservation laws. Theodoros KATSAOUNIS. Doron LEVY E COLE NORMALE SUPERIEURE A moded structured central scheme for 2D hyperbolc conservaton laws Theodoros KATSAOUNIS Doron LEVY LMENS - 98-30 Département de Mathématques et Informatque CNRS URA 762 A moded

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

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

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

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

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

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

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

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

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

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

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

Order of Accuracy Study of Unstructured Grid Finite Volume Upwind Schemes

Order of Accuracy Study of Unstructured Grid Finite Volume Upwind Schemes João Luz F. Azevedo et al. João Luz F. Azevedo joaoluz.azevedo@gmal.com Comando-Geral de Tecnologa Aeroespacal Insttuto de Aeronáutca e Espaço IAE 12228-903 São José dos Campos, SP, Brazl Luís F. Fguera

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

c 2009 Society for Industrial and Applied Mathematics

c 2009 Society for Industrial and Applied Mathematics SIAM J. MATRIX ANAL. APPL. Vol. 31, No. 3, pp. 1382 1411 c 2009 Socety for Industral and Appled Mathematcs SUPERFAST MULTIFRONTAL METHOD FOR LARGE STRUCTURED LINEAR SYSTEMS OF EQUATIONS JIANLIN XIA, SHIVKUMAR

More information

Abstract Ths paper ponts out an mportant source of necency n Smola and Scholkopf's Sequental Mnmal Optmzaton (SMO) algorthm for SVM regresson that s c

Abstract Ths paper ponts out an mportant source of necency n Smola and Scholkopf's Sequental Mnmal Optmzaton (SMO) algorthm for SVM regresson that s c Improvements to SMO Algorthm for SVM Regresson 1 S.K. Shevade S.S. Keerth C. Bhattacharyya & K.R.K. Murthy shrsh@csa.sc.ernet.n mpessk@guppy.mpe.nus.edu.sg cbchru@csa.sc.ernet.n murthy@csa.sc.ernet.n 1

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

Non-Split Restrained Dominating Set of an Interval Graph Using an Algorithm

Non-Split Restrained Dominating Set of an Interval Graph Using an Algorithm Internatonal Journal of Advancements n Research & Technology, Volume, Issue, July- ISS - on-splt Restraned Domnatng Set of an Interval Graph Usng an Algorthm ABSTRACT Dr.A.Sudhakaraah *, E. Gnana Deepka,

More information

Chapter 1. Comparison of an O(N ) and an O(N log N ) N -body solver. Abstract

Chapter 1. Comparison of an O(N ) and an O(N log N ) N -body solver. Abstract Chapter 1 Comparson of an O(N ) and an O(N log N ) N -body solver Gavn J. Prngle Abstract In ths paper we compare the performance characterstcs of two 3-dmensonal herarchcal N-body solvers an O(N) and

More information

An Accurate Evaluation of Integrals in Convex and Non convex Polygonal Domain by Twelve Node Quadrilateral Finite Element Method

An Accurate Evaluation of Integrals in Convex and Non convex Polygonal Domain by Twelve Node Quadrilateral Finite Element Method Internatonal Journal of Computatonal and Appled Mathematcs. ISSN 89-4966 Volume, Number (07), pp. 33-4 Research Inda Publcatons http://www.rpublcaton.com An Accurate Evaluaton of Integrals n Convex and

More information

an assocated logc allows the proof of safety and lveness propertes. The Unty model nvolves on the one hand a programmng language and, on the other han

an assocated logc allows the proof of safety and lveness propertes. The Unty model nvolves on the one hand a programmng language and, on the other han UNITY as a Tool for Desgn and Valdaton of a Data Replcaton System Phlppe Quennec Gerard Padou CENA IRIT-ENSEEIHT y Nnth Internatonal Conference on Systems Engneerng Unversty of Nevada, Las Vegas { 14-16

More information

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

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

More information

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

Computer models of motion: Iterative calculations

Computer models of motion: Iterative calculations Computer models o moton: Iteratve calculatons OBJECTIVES In ths actvty you wll learn how to: Create 3D box objects Update the poston o an object teratvely (repeatedly) to anmate ts moton Update the momentum

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

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

and NSF Engineering Research Center Abstract Generalized speedup is dened as parallel speed over sequential speed. In this paper

and NSF Engineering Research Center Abstract Generalized speedup is dened as parallel speed over sequential speed. In this paper Shared Vrtual Memory and Generalzed Speedup Xan-He Sun Janpng Zhu ICASE NSF Engneerng Research Center Mal Stop 132C Dept. of Math. and Stat. NASA Langley Research Center Msssspp State Unversty Hampton,

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

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

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

Query Clustering Using a Hybrid Query Similarity Measure

Query Clustering Using a Hybrid Query Similarity Measure Query clusterng usng a hybrd query smlarty measure Fu. L., Goh, D.H., & Foo, S. (2004). WSEAS Transacton on Computers, 3(3), 700-705. Query Clusterng Usng a Hybrd Query Smlarty Measure Ln Fu, Don Hoe-Lan

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

Multiblock method for database generation in finite element programs

Multiblock method for database generation in finite element programs Proc. of the 9th WSEAS Int. Conf. on Mathematcal Methods and Computatonal Technques n Electrcal Engneerng, Arcachon, October 13-15, 2007 53 Multblock method for database generaton n fnte element programs

More information

A gradient smoothing method (GSM) for fluid dynamics problems

A gradient smoothing method (GSM) for fluid dynamics problems INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN FLUIDS Int. J. Numer. Meth. Fluds 2008; 58:1101 1133 Publshed onlne 27 March 2008 n Wley InterScence (www.nterscence.wley.com)..1788 A gradent smoothng method

More information

Structured Grid Generation Via Constraint on Displacement of Internal Nodes

Structured Grid Generation Via Constraint on Displacement of Internal Nodes Internatonal Journal of Basc & Appled Scences IJBAS-IJENS Vol: 11 No: 4 79 Structured Grd Generaton Va Constrant on Dsplacement of Internal Nodes Al Ashrafzadeh, Razeh Jalalabad Abstract Structured grd

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

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

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

Virtual Machine Migration based on Trust Measurement of Computer Node

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

More information

IMPLEMENTATION OF UNSTRUCTURED GRID GMRES+LU-SGS METHOD ON SHARED-MEMORY, CACHE-BASED PARALLEL COMPUTERS

IMPLEMENTATION OF UNSTRUCTURED GRID GMRES+LU-SGS METHOD ON SHARED-MEMORY, CACHE-BASED PARALLEL COMPUTERS AIAA-97 IMPLEMENTATION OF UNSTRUCTURED GRID GMRES+LU-SGS METHOD ON SHARED-MEMORY, CACHE-BASED PARALLEL COMPUTERS Dmtr Sharov, Hong Luo, Joseph D. Baum Scence Applcatons Internatonal Corporaton 7 Goodrdge

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

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

Newton-Raphson division module via truncated multipliers

Newton-Raphson division module via truncated multipliers Newton-Raphson dvson module va truncated multplers Alexandar Tzakov Department of Electrcal and Computer Engneerng Illnos Insttute of Technology Chcago,IL 60616, USA Abstract Reducton n area and power

More information

Exercises (Part 4) Introduction to R UCLA/CCPR. John Fox, February 2005

Exercises (Part 4) Introduction to R UCLA/CCPR. John Fox, February 2005 Exercses (Part 4) Introducton to R UCLA/CCPR John Fox, February 2005 1. A challengng problem: Iterated weghted least squares (IWLS) s a standard method of fttng generalzed lnear models to data. As descrbed

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

Discontinuous Galerkin methods for flow and transport problems in porous media

Discontinuous Galerkin methods for flow and transport problems in porous media T COMMUNICATIONS IN NUMERICA METHODS IN ENGINEERING Commun. Numer. Meth. Engng 2; :1 6 [Verson: 2/3/22 v1.] Dscontnuous Galerkn methods for flow and transport problems n porous meda Béatrve Rvère and Mary

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

Some material adapted from Mohamed Younis, UMBC CMSC 611 Spr 2003 course slides Some material adapted from Hennessy & Patterson / 2003 Elsevier

Some material adapted from Mohamed Younis, UMBC CMSC 611 Spr 2003 course slides Some material adapted from Hennessy & Patterson / 2003 Elsevier Some materal adapted from Mohamed Youns, UMBC CMSC 611 Spr 2003 course sldes Some materal adapted from Hennessy & Patterson / 2003 Elsever Scence Performance = 1 Executon tme Speedup = Performance (B)

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

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

Overview. Basic Setup [9] Motivation and Tasks. Modularization 2008/2/20 IMPROVED COVERAGE CONTROL USING ONLY LOCAL INFORMATION

Overview. Basic Setup [9] Motivation and Tasks. Modularization 2008/2/20 IMPROVED COVERAGE CONTROL USING ONLY LOCAL INFORMATION Overvew 2 IMPROVED COVERAGE CONTROL USING ONLY LOCAL INFORMATION Introducton Mult- Smulator MASIM Theoretcal Work and Smulaton Results Concluson Jay Wagenpfel, Adran Trachte Motvaton and Tasks Basc Setup

More information

In the planar case, one possibility to create a high quality. curve that interpolates a given set of points is to use a clothoid spline,

In the planar case, one possibility to create a high quality. curve that interpolates a given set of points is to use a clothoid spline, Dscrete Farng of Curves and Surfaces Based on Lnear Curvature Dstrbuton R. Schneder and L. Kobbelt Abstract. In the planar case, one possblty to create a hgh qualty curve that nterpolates a gven set of

More information

An inverse problem solution for post-processing of PIV data

An inverse problem solution for post-processing of PIV data An nverse problem soluton for post-processng of PIV data Wt Strycznewcz 1,* 1 Appled Aerodynamcs Laboratory, Insttute of Avaton, Warsaw, Poland *correspondng author: wt.strycznewcz@lot.edu.pl Abstract

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

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

Construction of ROBDDs. area. that such graphs, under some conditions, can be easily manipulated.

Construction of ROBDDs. area. that such graphs, under some conditions, can be easily manipulated. A Study of Composton Schemes for Mxed Apply/Compose Based Constructon of s A Narayan 1 S P Khatr 1 J Jan 2 M Fujta 2 R K Brayton 1 A Sangovann-Vncentell 1 Abstract Reduced Ordered Bnary Decson Dagrams

More information

Communication-Minimal Partitioning and Data Alignment for Af"ne Nested Loops

Communication-Minimal Partitioning and Data Alignment for Afne Nested Loops Communcaton-Mnmal Parttonng and Data Algnment for Af"ne Nested Loops HYUK-JAE LEE 1 AND JOSÉ A. B. FORTES 2 1 Department of Computer Scence, Lousana Tech Unversty, Ruston, LA 71272, USA 2 School of Electrcal

More information

Polyhedral Compilation Foundations

Polyhedral Compilation Foundations Polyhedral Complaton Foundatons Lous-Noël Pouchet pouchet@cse.oho-state.edu Dept. of Computer Scence and Engneerng, the Oho State Unversty Feb 8, 200 888., Class # Introducton: Polyhedral Complaton Foundatons

More information

Topology Design using LS-TaSC Version 2 and LS-DYNA

Topology Design using LS-TaSC Version 2 and LS-DYNA Topology Desgn usng LS-TaSC Verson 2 and LS-DYNA Wllem Roux Lvermore Software Technology Corporaton, Lvermore, CA, USA Abstract Ths paper gves an overvew of LS-TaSC verson 2, a topology optmzaton tool

More information

Differential formulation of discontinuous Galerkin and related methods for compressible Euler and Navier-Stokes equations

Differential formulation of discontinuous Galerkin and related methods for compressible Euler and Navier-Stokes equations Graduate Theses and Dssertatons Graduate College 2011 Dfferental formulaton of dscontnuous Galerkn and related methods for compressble Euler and Naver-Stokes equatons Hayang Gao Iowa State Unversty Follow

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

APPLICATION OF A COMPUTATIONALLY EFFICIENT GEOSTATISTICAL APPROACH TO CHARACTERIZING VARIABLY SPACED WATER-TABLE DATA

APPLICATION OF A COMPUTATIONALLY EFFICIENT GEOSTATISTICAL APPROACH TO CHARACTERIZING VARIABLY SPACED WATER-TABLE DATA RFr"W/FZD JAN 2 4 1995 OST control # 1385 John J Q U ~ M Argonne Natonal Laboratory Argonne, L 60439 Tel: 708-252-5357, Fax: 708-252-3 611 APPLCATON OF A COMPUTATONALLY EFFCENT GEOSTATSTCAL APPROACH TO

More information