RESTRICTED ADDITIVE SCHWARZ METHOD WITH HARMONIC OVERLAP

Size: px
Start display at page:

Download "RESTRICTED ADDITIVE SCHWARZ METHOD WITH HARMONIC OVERLAP"

Transcription

1 RESTRICTED ADDITIVE SCHWARZ METHOD WITH HARMONIC OVERLAP XIAO-CHUAN CAI, MAKSYMILIAN DRYJA, AND MARCUS SARKIS Abstract. In ths paper, we ntroduce a new Schwarz precondtoner and wth a new coarse space. We construct the precondtoner by groupng together, n one precondtoner, features from the addtve overlappng Schwarz methods, from teratve substructurng methods and from a class of restrcted addtve Schwarz methods. The precondtoner s symmetrc and consdered as a symmetrzed verson of restrcted addtve Schwarz precondtoners. We also enhance the precondtoner wth a new coarse space whch s smple, easly parallelzable, has smaller stencl, and has one degree of freedom per substructure and can be used n problems wth unstructured meshes. We study the spectral bounds for the method and dscuss the several advantages of ths precondtoners. Numercal results theory wll be provded. Key words. Schwarz precondtoner, fnte element, overlappng doman decomposton, ellptc equatons, coarse space AMS(MOS) subject classfcatons. 65N30, 65F0. Introducton. Through out ths paper, C and c, are postve generc constants that are ndependent of any of the mesh parameters and the number of subdomans. All the domans and subdomans are assumed to be open,.e., boundares are not ncluded n ts defnton. 2. Dscretzaton. Fnd u H 0 (Ω), such that (2.) where a(u, v) = Ω a(u, v) = f(v) v H0 (Ω), u v dx and f(v) = fv dx for f L 2 (Ω). Ω For smplcty, let Ω be a bounded polygonal regon n R 2 wth a dameter of sze O(). The extenson of the results to R 3 can be carred out easly by usng the theory developed here n ths paper and the well-known three-dmensonal addtve Schwarz technques; [5, 6, 7, 4]. We ntroduce a trangulaton T h (Ω) whch s shape regular and for smplcty quas-unform of sze h. We then defne a fnte element space V(Ω), whch we denote also by V, whch s the space of contnuous pecewse lnear functons assocated wth the trangulaton T h (Ω) and whch vanshes on Ω, the boundary of Ω. We are nterested n solvng the followng dscrete problem assocated to (2.): Fnd u V such that (2.2) a(u, v) = f(v), v V. 3. Notatons and Defnton of Subspaces. Let W be the set of the n nteror nodes of T h (Ω). We assume that the graph parttonng has been appled and has resulted n N nonoverlappng subsets W 0, =,..., N whose unon s W. We defne the overlappng partton of W as follows. Let {W } be the one-overlap partton of Dept. of Comp. Sc., Unv. of Colorado, Boulder, CO 80309, ca@cs.colorado.edu. Faculty of Math. Info. and Mech., Warsaw Unv., Warsaw, dryja@mmuw.edu.pl. Mathematcal Scences Department, Worcester Polytechnc Insttute, Worcester, MA msarks@wp.edu

2 W, where W W 0 s obtaned by ncludng all the mmedate neghborng vertces of the vertces n W 0. Usng the dea recursvely, we can defne a δ-overlap partton of W, W = N W δ, = where W δ W 0 wth δ levels of overlaps wth ts neghborng subdomans. Here δ 0 s an nteger, and δh s the lenght of the extenson. We defne regons Ω R by the unon of all the elements of T h (Ω) whch have all the three vertces on W 0 Ω. We denote H the representatve sze of the subregon Ω R. We note that we can always defne a subregon nduced by a set of nodes. Assume for nstance that a set Z s a subset of W WB 0. Here W B 0 s the set of boundary nodes. We denote nduced subregon of Z by Ω(Z) defned as the unon of Z, unon the open elements (trangles) of T h (Ω) whch have at least one vertex of Z, and unon of the open edges of these trangles whch has at least one endpont as a vertex of Z. we note that Ω(Z) s always an open regon f Z s a subset of W. We defne the extended regon Ω δ by Ω(W δ). We ntroduce the space V V H0 (Ω δ ) extended by zero outsde Ω\Ωδ. It s easy to check that we can decompose V as V = V + V V N. Ths decomposton wll be used n defnng the regular addtve Schwarz algorthm wthout a coarse space [3, 0]. In order to defne the new method,.e. the restrcted addtve Schwarz method wth harmonc overlap wth the new coarse space, we have to ntroduce an mportant dea of how we deal wth boundary Drchlet data. We note that we do not need to examne dfferent geometrc cases as they appear n the studes of teratve substructurng methods. There, t s needed to consder geometrcal cases such as f a subdoman touches the boundary n one pont or a whole edge; or f a subdoman should have a degrees of freedoms assocate to the coarse space. Ths consderaton also appears n the regular addtve Schwarz method wth an exotc coarse spaces [2, 5,,?, 7]. Let Ω 0 B = Ω(W B 0 ),.e. one layer of elements near the boundary Ω. We note that Ω belongs to Ω B. We also defne W ˆδ B... WB W B 0 wth ˆδ levels of extenson by addng recursvely neghborng vertces. And defne Ωˆδ B = Ω(W ˆδ B ). We note that ˆδ s gong to be same order as δ. The reason to choose ˆδ possbly dfferent to δ s because the overlap between Ωˆδ B and a Ω δ next to the boundary s (ˆδ, whle the overlap between two floatng subdomans are (2δ. Therefore, choosng ˆδ = 2δ may be more approprate than choosng ˆδ = δ. We wll see ths clearly n the analyss. We treatment of mxed Drchlet and Neumann boundary condtons s also trval. The only thng we need to do s to move de boundares assocated to Neumann part from WB 0 to W. We wll not treat ths case n the paper snce no new deas are necessary. We note that the coarse space that we ntroduce n ths paper, one degree of freedom per subdoman s used ndependent f t s a Drchlet or Neumann data, or how the subdoman touches the boundary. Let Γ = Ω δ \ Ω,.e., the part of the boundary of Ωδ that does belong to the Drchlet boundary part. Also let Γ B = Ωˆδ B \ Ω. We defne the nterface overlappng boundary Γ as the unon of all the Γ and Γ B,.e. Γ = Γ B ( N = Γ ). We then defne the followng subset of nodes of W : 2

3 W Γ W Γ (the nterface nodes) W Γ W Γ W δ (the local nterface nodes) W,n Γ W Γ W 0 (the local nternal nterface nodes) W,cut Γ W Γ\W,n Γ (the local cut nterface nodes) W,ovl δ (W δ\w Γ) (( j W j δ) W ˆδ B ) (the local overlappng nodes) W,non δ W δ\(w Γ W,ovl δ ) (the local nonoverlappng nodes) W,n δ W,non W,n Γ (the nternal nodes) We remark that the sets W,n Γ form a nonoverlappng decomposton of W Γ and t wll play a very mportant role n the defnton of the new Schwarz methods and n the new coarse space ntroduced later n ths paper. Let x k W be a mesh pont and φ xk V the fnte element bass functon assocated wth x k,.e. φ xk (x k ) =, and φ xk (x j ) = 0, j k. We say u V s dscrete harmonc at x k f a(u, φ xk ) = 0. If u s dscrete harmonc at a set of nodal ponts Z, we say u s dscrete harmonc n Ω(Z). We defne Ṽ as a subspace of V consstng of functons that vansh on the cutng nodes W,cut Γ and dscrete harmonc at the nodes W,ovl δ. We defne Ṽ as a subspace of V defned as Ṽ = Ṽ + Ṽ2 + + ṼN. It s easy to see that the sum above s a drect sum. We defne P : Ṽ Ṽ to be the projecton operators such that, for any u Ṽ (3.) a( P u, v) = a(u, v), v Ṽ. The restrcted addtve Schwarz wth harmonc overlap method (RASH) s defne as P = P. We also defne P : V V by = (3.2) a(p u, v) = a(u, v), v V. The regular addtve overlappng Schwarz method (AS) s defned as P = P. = Equvalentelly, we can defne P and P n matrx notatons. Let us defne W δ W δ\w,cut Γ, and let Ω( W δ) be the nduced doman. It s easy to see that Ω( W δ ) s the same as Ω δ but wth cuts. We denote Ω( W δ) by Ω δ. We have then Ṽ = V H0 ( Ω δ ) and dscrete harmonc on Ω(W,ovl δ ). Let R δ : W W δ be restrcton operator assocated to the the regular pontwse nterpolaton from V to Ṽ. And R δ : W W δ the restrcton operator from V to V. 3

4 Then P u, for u Ṽ n matrx notaton s defned as follows: and P u, for u V s defned as where P u = P u = ( R δ ) T (Ãδ ) Rδ Au = (R δ ) T (A δ ) R δ Au = Ã δ = R δ A( R δ ) T A δ = R δ A(R δ ) T. 4. A New Coarse Space. We next ntroduce a new coarse space Ṽ0. The assocated coarse problem wll be then added to P. Ṽ 0 s a subspace of Ṽ and t s defned by the followng coarse bass functons { f node φ xk W,n δ 0(x k ) = 0 f node x k W \ W δ and φ 0 s dscrete harmonc at x k W,ovl δ. We note that support of φ 0 s n Ω δ and therefore ts constructon can be performed n parallel and wth no communcaton. We remark also that the coarse space Ṽ 0 can also be used as a coarse space for the regular addtve Schwarz method snce Ṽ 0 s a subspace of Ṽ, and Ṽ a subspace of V 0. An nterpolaton-lke operator I 0 = N = I 0, where the I0 : V Ṽ s defned as follows: (I 0u)(x) = ū φ 0(x). Here ū = Ω δ udx. Ω δ The value ū s the average of u on the extended regon Ω δ. Here Ωδ s the area of the regon Ω δ. We have then (I 0 u)(x) = = ( Ω δ Ω δ udx)φ 0(x). We defne the coarse space Ṽ0 as the range of I 0, or equvalently, as a lnear combnatons of the φ 0. We ntroduce P 0 : V Ṽ0 as the operator such that, for any u V, (4.) a( P 0 u, v) = a(u, v), v Ṽ0 4

5 Let R 0 T : Ṽ0 V be the standard embedded prolongaton operator. The coarse projecton operator P 0 n matrx notaton s defned by P 0 = R 0 T Ã 0 R 0 A where Ã0 = R 0 A R 0 T. We remark that Ã0(, j) = 0 f domans and j are not neghbors. We note that Ã0 s more sparse than coarse spaces matrces that appear n other methods such as Neumann-Neumann [9] and FETI [8] precondtoners. Another mportant feature of ths coarse space problem s that the computaton of rght hand sde,.e. the nner product (φ 0, r) can be computed nsde Ω δ, therefore ts parallelzaton s straghtforward to perform and no communcatons are necessary. The restrcted addtve Schwarz method wth harmonc overlap (RASH) wth coarse space s then defne as P = P 0 + P. 5. Theoretcal Analyss. In order to analyze the method, we ntroduce a specal functon, denoted by ϑ whch ts energy sem-norm can be easly estmated and control the energy sem-norm of φ 0. The ϑ s used only for analyss purposes. We note that ϑ and φ 0 are dentcal ex cept at the nodes W δ,ovl. In next Lemma we study the energy sem-norm of each coarse bass functon. Lemma 5.. For any φ 0 and δgeq0 and ˆδ = O(δ). Then φ 0 2 H (Ω) = H C( + log((δ ) + h (2δ ) Proof. Consder frst the case n whch Ω δ s floatng square subdoman. Later we extend the cases n whch Ω δ s near the boundary. To smplfy our arguments, we assume n the analyss of ths and the remanng lemmas of the paper that the Ω δ and ts neghborng extended subdomans Ω δ j are squares of the same sze,.e. sdes length equals H + 2(δ. Ths assumpton s equvalent to that Ω R has sze H and δ level of overlap s appled. And also assume the overlap s not too large; for the analyss gven below (δ)h no larger than H/4 s enough. We note that too large overlap s not of practcal mportance. We note the our technques can be modfed to consder larger overlaps and more complex subdomans. We next partton Ω δ nto subregons n order to defne ϑ. ϑ s defned as a pecewse lnear functon of V wth support on Ω δ. It s equals to φ 0 on W \W,ovl δ. On the overlap regon W,ovl δ we defne ϑ carefully so that we can control ts energy sem-norm. Because φ 0 s dscrete harmonc on W,ovl δ, we have a(φ 0, φ 0) a(ϑ, ϑ ). Before we start defnng ϑ we need to ntroduce some notaton to characterze each pece of the subdoman Ω δ. We splt Ωδ nto subregons of four types: ΩI, Ωδδ, Ω δh, and Ω δ δ. The frst subregon, ΩI, s defned as Ω( W,non ). Ths subregon s therefore a square wth sdes of sze H 2δh. The second subregon, Ω δδ, s the place where Ω δ overlaps smulatneously three neghbors Ωδ j. Ωδδ therefore represents the 5

6 unon of the four corner peces of Ω δ,.e. four squares wth sdes of sze (2δ. The places where Ω δ overlaps only one neghbor are four rectangles wth sdes of szes H (2δ)h and (2δ. We then partton each of these four rectangles nto three subrectangles,.e. two of them are of Ω δ δ type and one of them of Ω δh type. For nstance, wthout lost of generalty assume we want the partton the ntersecton of Ω δ and and ts rght neghbor Ωδ j, excludng the the corner parts. Therefore regon we want to partton s a rectangle wth edges szes (2δ +)h n x drecton and H (2δ)h n y drecton. The partton of ths rectangles gves two subrectangles of Ω δ δ type wth dmensons 2(δ+)h (δ)h and whch each one has a whole edge n common wth a square of Ω δδ type. We denote them as transton subregons because t s placed between a corner type subregon Ω δδ and a face type subregon Ω δh. The Ω δh face type subregons are the subrectangles that are placed between the two subrectangles of Ω δ δ type. Ω δh face type regons have sdes of szes (2δ and H (4δ)h. On Ω I and we defne ϑ to be equals one,.e. equals to φ 0. Therefore φ 0 2 H (I) = ϑ 2 H (I) = 0. The most nterestng part s how to defne ϑ on Ω(W,ovl δ ). Bascally, Ω(W,ovl δ ) s the unon of corner, transton and face type regons defned above. We next defne ϑ n these regons and estmate ts energy sem-norm. The defnton of ϑ s gven by defnng ϑ (x) for all nodes x. We frst examne the case of a corner type square. Assume Q s such a square wth coordnates of ts vertces gven by V = [a, b], V 2 = [a + (2δ, b], V 3 = [a, b + (2δ ], and V 4 = [a + (2δ, b + (2δ ]. Assume also that V, V 2, and V 4 belong to Ω δ. In other words, the Q s placed on the southeast corner part of Ω δ. We then also ntroduce the square Q, wth vertces V 3 = [a, b + (2δ ], Ṽ = [a, b+(δ)h], Ṽ2 = [a+(δ +)h, b+(δ)h], and Ṽ4 = [a+(δ +)h, b+(δ +)h]. Note that Q s contaned n Q wth the vertex V 3 as the common vertex. We next defne ϑ on Q. We set ϑ (V 3 ) =, ϑ (Ṽ) = 0, ϑ (Ṽ2) = 0, ϑ (Ṽ4) = 0. The remanng nodes x on the edges ṼṼ2 and Ṽ2Ṽ4 we set ϑ (x) = 0, and on the edge V 3 Ṽ and V 3 Ṽ 4 we set ϑ (x) =. For nodes on Q\ Q we set ϑ (x) = 0. It remans only to defne ϑ (x) for nodes x n the nteror of Q. To defne ϑ there we use well-known cut functons technques such as those ntroduced n Lemma 4.4 of [5] but appled for two-dmensonal square regons. The most mportant observaton of the constructon of ϑ nsde Q s that ϑ (x) C/r near Ṽ or Ṽ4. Here r s dstance of x to Ṽ or Ṽ 4. Therefore, we obtan (see [5]) ϑ 2 H (Q) = ϑ 2 H ( Q) C( + log((δ )) h Snce nsde Ω δδ there four of those squares we obtan ϑ 2 H (Ω δδ ) C( + log((δ )) h We next consder the case of a transton type rectangle, denoted by T, whch we assume the coordnates of ts vertces are gven by V 3 = [a, b + (2δ ], V 4 = [a+(2δ+)h, b+(2δ+)h], V 5 = [a, b+(3δ+)h], and V 6 = [a+(2δ+)h, b+(3δ+)h]. Note that T stands on the top of the square Q ntroduced above and has the edge V 3 V 4 n common. We defne ϑ over the edge V 3 V 4 to be equals φ 0. Over the edge 6

7 V 3 V 5, we set ϑ =. Over edge V 4 V 6, we set ϑ = 0. And over edge V 5 V 6 we let ϑ to decreases lnearly from value to 0. It remans to defne ϑ nsde T. Let us defne the nodes V l = [a + (δ)h, b + (2δ ] and V r = [a + (δ, b + (2δ ]. These nodes are exactly the places over the sde V 3 V 4 where φ 0 jumps from to 0. On the trangle V 3 V l V 5 we set ϑ =. On the trangle V r V 4 V 6 we set ϑ = 0. On the regon V l V r V 6 V 5, we defne ϑ to decreases lnearly n the x drecton from value to 0. We note that next to the nodes V l V r, ϑ has a sngular behavor smlar to ϑ (x) C/r where r s the dstance from x to the lne V l V r. Agan, we also obtan ϑ 2 H (T ) C( + log((δ )). h Snce nsde of Ω δ δ contans eght rectangles of T type we obtan ϑ 2 H (Ω δ δ C( + log((δ )). ) h Fnally we consder the rectangle of face type, denoted by R, assumng that the coordnates of ts vertces are gven by V 5 = [a, b + (3δ ], V 6 = [a + (2δ, b + (3δ ], V 7 = [a, b + H (δ )h], and V 8 = [a + (2δ, b + H (δ )h]. Note that R s on the top of the rectangle T defned above and ts heght s H (4δ)h. The vertces V 6 and V 8 are the vertces that belong to Ω δ. We defne ϑ (x) = f x s on the edge V 5 V 7, and equals zero f x n on the edge V 6 V 8, and lnear n x drecton on the remanng ponts. We obtan then ϑ 2 H (R) H (4δ)h (2δ. Snce there are four of those rectangles nsde Ω δh, we obtan ϑ H (Ω δh ) C H (4δ)h (2δ C H (2δ. Puttng all every peces ϑ together, we can see that ϑ V and t s equals to φ 0 on W \W δ,non. The lemma follows by usng that φ 0 at the φ 0 H (Ω δ ) ϑ H (Ω δ ) C( + log( (δ H ) + h (2δ ) by usng all the estmates of subregons of four types dscussed. For the cases n whch Ω s near the boundary Ω the analyss does not change much. The analyss s smlar and we deal stll wth the four types subregons dscussed above. The basc dfference s that the sze of the T, R, I, and Q change slghtly from before, however stll at the same order of magntude ˆδ = O(δ). And the proof follows. Lemma 5.2. For any u V we have I 0 u 2 H (Ω) H C( + log((δ ) + h (2δ ) u 2 H (Ω). Proof. Let I 0 u = N = ūφ 0, where ū = Ω δ Ω δ udx. 7

8 Let c be a constant. By usng Cauchy-Schwarz nequalty we have ū c = Ω δ (u c)dx C H u c L 2 (Ω δ ). Ω δ Assume that Ω δ j s neghbor to Ωδ,.e. Ωδ j overlaps Ωδ j. Usng a trangular nequalty we obtan ū ū j ū c + ū j c C H u c L 2 (Ω δ Ωδ j ). And usng a Bramble-Hlbert argument we obtan (5.) ū ū j C nf c H u c L 2 (Ω δ Ωδ j ) C nf c H u c L 2 (Ω δ,ext ) C u H (Ω δ,ext ). Here Ω δ,ext s the unon of Ω δ and ts all overlappng neghbors extended regons Ωδ k. Note that we had used that Ω δ,ext has a regular shape and has sze O(H) to use a Fredrch nequalty. We now are ready to evaluate the energy norm of I 0 u on Ω δ. We use strongly the fact that φ 0 =, on Ω δ. j= On Ω I, φj 0 vanshes when j and equals one when j =. Therefore, I 0 u 2 H (I) = ū 2 H (I) = 0. We next evaluate I 0 u on a corner type square Q. Wthout less of generalty, ths corner s the southeast corner of Ω δ, and overlaps wth Ωδ, Ω δ 2, and Ω δ 3,.e. the east, south, and southeast regons. Then We next use that to obtan and by a trangular nequalty I 0 u 2 H (Q) = ū φ 0 + φ j= 3 j= φ k 0 = on Q ū k φ k 0 2 H (Q). 3 I 0 u 2 H (Q) = (ū k ū )φ k 0 2 H (Q), k= k= 3 3 I 0 u 2 H (Q) C (ū k ū )φ k 0 2 H (Q) C ū k ū φ k 0 2 H (Q). 8 k=

9 Usng (5.) and same arguments n the proof of Lemma 5. we obtan I 0 u 2 H (Q) C( + log((δ )) u h H (Ω δ,ext ). Usng smlar arguments, we obtan for a rectangle of face type R and transton type T, I 0 u 2 H (R) C( + H (2δ ) u 2 H (Ω δ,ext ) I 0 u 2 H (T ) Summng all the contrbutons, we have I 0 u 2 H (Ω δ C( + log((δ )) u 2 h. H (Ω δ,ext ) H ) C( + log((δ ) + h (2δ ) u 2. H (Ω δ,ext ) Agan, the analyss s smlar for the cases n whch Ω s near the boundary. The bounds wll be of the same order as before f ˆδ = O(δ). And the proof follows. Theorem 5.. There exsts a constant C > 0, ndependent of h and the number of subdomans, such that for any u Ṽ, there exst u Ṽ, such that and u = u, =0 (5.2) u 2 H (Ω) =0 C[( + log((δ ))( + log( H h h )) + H (2δ ] u 2 H (Ω) Recall that δ s a nonnegatve nteger ndcatng the number of overlap and δh s the actual overlappng sze. Proof. Let the local functons v Ṽ be defned as { u(xk ) f f node x k W,n δ v (x k ) = 0 f f node x k W \ W δ and v s dscrete harmonc at the x k W δ,ovl Let u 0 Ṽ0 be defned as u 0 = I 0 u = I0u = ū φ 0. = = The decomposton u = N =0 u s defned as follows u = v ū φ 0, =,..., N 9

10 u 0 = ū φ 0. = The next step s to bound N =0 u 2 H (Ω). For boundng u 0 2 )H (Ω) we use Lemma 5.2,.e u 0 2 H (Ω) H C( + log((δ ) + h (2δ ) u 2 H (Ω). To bound u 2 H (Ω), =,..., N, we use the ϑ, =,..., N ntroduced n Lemma 5.. We ntroduce ũ Ṽ as follows ũ (x) = I h (ϑ (x)(u(x) ū )). Note that ũ (x) s equal to u (x) n the whole Ω but Ω(W δ,ovl ). On Ω(W δ,ovl ) u s dscrete harmonc. Therefore, we have u 2 H (Ω) ũ 2 H (Ω). So, the next step s to bound ũ 2 H (Ω). Let K be an element of Ωδ denote w = u ū. We obtan ũ 2 H (K) = I h(ϑ w ) 2 H (K) 2 ϑ w 2 H (K) + 2 I h(( ϑ ϑ )w ) 2 H (K). and let us Here, ϑ s the average of ϑ on K, and I h s the standard pontwse nterpolator. To bound the frst part we use that ϑ, to obtan ϑ w 2 H (K) ϑ (u ū ) 2 H (K) u ū 2 H (K) C u 2 H (K). The last nequalty comes from the fact that ū s a constant. For the second part we frst use an nverse nequalty to have I h (( ϑ ϑ )w ) 2 H (K) C h 2 I h(( ϑ ϑ )w ) 2 L 2 (K). We next splt the regon Ω δ n the 4 peces that we had already ntroduced n Lemma 5.,.e. Ω I, ΩδH, Ω δ δ and Ω δδ. For K on Ω I, ϑ ϑ s dentcally to zero, and therefore, I h (( ϑ ϑ )w ) 2 L 2 (K) vanshes. For K on Ω δh we use that ϑ ϑ 2 L (K) C( h (2δ )2 to obtan h 2 I h(( ϑ ϑ )w ) 2 L 2 (K) C ((2δ ) 2 w 2 L 2 (K). And we use small overlap technques Dryja and Wdlund [6] to obtan ((2δ ) 2 w 2 L 2 (Ω H δ ) C( + H (2δ w 2 H (Ω δ ) + H((2δ ) w 2 L 2 (Ω δ )) 0

11 We use that w 2 H (Ω δ ) = u 2 and a Fredrch nequalty H (Ω δ ) Puttng everythng together, we obtan K Ω δδ w 2 L 2 (Ω δ ) CH2 u 2 H (Ω δ ). (2δ ) 2 w 2 L 2 (Ω H δ ) C( + H (2δ ) u 2 H (Ω δ ) For K on Ω δδ, we use smlar arguments as n Dryja, Smth and Wdlund [5] to obtan ((δ ) 2 I h(( ϑ ϑ )w 2 L 2 (K) r 2 w 2 L 2 (K) K Ω δδ K Ω δδ where ch r C((δ ) s the dstance to those cut peces. We have used here that ϑ has the sngular behavor C/r on Ω δδ. We have then C(δ+)h r 2 w 2 L 2 (K) C r 2 r w 2 L (Ω δδ)dθdr ch θ C( + log( (δ h )) w 2 L (Ω δδ ) C[( + log((δ ))( + log( H h h ))] u 2 Ω. δ For the last nequalty we had used a well-known result (see Bramble []) u ū 2 L (Ω δδ ) u ū L (Ω δ ) C(+log( H h )) u ū 2 H (Ω δ )andusngthat ū s the average of u on Ω δ,.e. a Fredrch nequalty u ū 2 H (Ω δ ) C( + log( H h )) u 2 H (Ω δ ) Usng smlar arguments, we also obtan a smlar bound fot the Ω δ δ. Agan, the analyss s smlar for the cases n whch Ω s near the boundary. The bounds wll be of the same order as before f ˆδ = O(δ). And the proof follows. 6. Numercal experments. In ths secton, we present some numercal results. We consder the Posson s equaton on the unt square wth zero Drchlet boundary condton. We consder the PCG, wth RASH or AS as precondtoner. In order to apply the RASH/PCG method, t s necessary to force the soluton to belong to Ṽ. To do so, an addtonal pre computaton s needed to reduce to ths case. Ths s done by applyng one precondtoned Rchardson teraton gven by w = ( R δ ) T (Ãδ ) R0 f. = We note that u = u w Ṽ, and therefore we can apply the RASH/PCG. The AS/CG s the classcal addtve Schwarz precondtoned CG as descrbed n [?]. We shall use a fne mesh. The stoppng condton for CG s to reduce the ntal

12 Table 6. RASH (AS) performance results wthout coarse solver for solvng the Posson s equaton on a mesh decomposed nto 2 2 = 4 subdomans wth overlap = ovlp. ovlp Iter Cond Max Mn 0 42 (42) 29.(29.).98 (.98) (0.054) 24+ (28) 48.4 (86.3).94 (4.00) (0.0464) (23) 33.3 (5.8).9 (4.00) (0.0773) 3 8+ (20) 27.2 (37.0).89 (4.00) (0.08) Table 6.2 RASH (AS) performance results wthout coarse solver for solvng the Posson s equaton on a 32 DOM 32 DOM mesh decomposed nto DOM DOM subdomans wth overlap =. DOM DOM Iter Cond Max Mn (20) 26.8 (43.7).89 (4.00) (0.096) (42) 86.9 (45.).95 (4.00) (0.0276) (78) 328. (550.).97 (4.00) (0.0073) (56) 295 (268.).98 (4.00) 0.005(0.008) resdual by a factor 0 6. The teraton counts, condton numbers, maxmum and mnmum egenvalues are summerzed n Tables 6, 6, and 6. From Tables 6 and 6, t s clear that RASH/PCG s always better than the classcal AS/PCG n terms of the teraton counts and condton numbers for the case that no coarse space s used. Ths s an mportant result snce t s easy to modfy a (parallel) code wth AS/PCG to RASH/PCG. We also predct that the RASH/CG would be even better than AS/CG n a parallel computer wth dstrbuted memory snce much less communcatons are requred. Also the local solvers are slghtly cheaper snce the local solvers has a slghtly smaller number of unknowns than for the regular AS. We also tested our new coarse space for the RASH/PCG. Table 6 shows that ths coarse space s very effectve when the number of subdomans gets very large. Ths s encouragng and we predct that ths coarse can be very promsng for parallel mplementaton snce t s easy to program, t requres fewer communcatons and computatons than the exstng coarse spaces. The results are for the case we use u(x, y) = e 5(x+y) sn(πx) sn(πy) as the exact soluton. We note that dfferent rght hand sdes and soluton functons were tested; the results follow the same patterns as below. REFERENCES [] J. Bramble, A second order fnte dfference analogue of the frst bharmonc boundary value problem, Numer. Math., 9, 966, pp [2] X.-C. Ca e M. Sarks, A restrcted addtve Schwarz precondtoner for general sparse lnear systems, SIAM J. Sc. Comput., 2 (999), pp [3] M. Dryja and O. Wdlund, An addtve varant of the Schwarz alternatng method for the case of many subregons, Techncal Report 339, also Ultracomputer Note 3, Department of Computer Scence, Courant Insttute, 987 [4] M. Dryja, M. Sarks, and O. Wdlund, Multlevel Schwarz methods for ellptc problems wth dscontnuous coeffcents n three dmensons, Numer. Math., Vol. 72, 996, pp [5] M. Dryja, B. Smth, and O. Wdlund, Schwarz analyss of teratve substructurng algo- 2

13 Table 6.3 RASH performance results wth a coarse solver for solvng the Posson s equaton on a 32 DOM 32 DOM mesh decomposed nto DOM DOM subdomans wth overlap = (2). DOM DOM Iter Cond Max Mn (6+) 30.2 (22.3) 2.67 (2.63) (0.79) (33+) 46.6 (32.6) 2.9 (2.88) (0.0886) (44+) 57.7 (4.0) 2.96 (2.93) (0.077) (46+) 6.4 (44.0) 2.96 (2.94) (0.0669) rthms for ellptc problems n three dmensons, SIAM J. Numer. Anal., Vol. 3(6), 994, pp [6] M. Dryja and O. Wdlund, Doman decomposton algorthms wth small overlap, SIAM J. Sc. Comp., 5, 994, pp [7] M. Dryja and O. Wdlund, Schwarz Methods of Neumann-Neumann type for threedmensonal ellptc fnte elements problems, Comm. Pure Appl. Math, Vol. 48, 995, pp [8] C. Farhat and F. Roux, A Method of Fnte Element Tearng and Interconnectng and ts Parallel Soluton Algorthm, Int. J. Numer. Mech. Engrg., Vol. 32, 99, pp [9] J. Mandel, Balancng Doman Decomposton, Comm. Numer. Meth. Eng., Vol. 9, 993, pp [0] A. Matsokn and S. Nepomnyaschkh, A Schwarz alternatng method n a subspace, Sovet Mathematcs, Vol. 29(0), 985, pp [] M. Sarks, Nonstandard coarse spaces and Schwarz methods for ellptc problems wth dscontnuous coeffcents usng non-conformng elements, Numersche Mathematk, Vol 77, 997, [2] O. Wdlund, Exotc Coarse Spaces for Schwarz Methods for Lower Order and Spectral Fnte Element, Contemporary Mathematcs, Seventh Internatonal Conference of Doman Decomposton Methods n Scentfc and Engneerng Computng, Davd E. Keyes and Jnchao Xu, Vol 80, 994,

Additive Schwarz Method for DG Discretization of Anisotropic Elliptic Problems

Additive Schwarz Method for DG Discretization of Anisotropic Elliptic Problems Addtve Schwarz Method for DG Dscretzaton of Ansotropc Ellptc Problems Maksymlan Dryja 1, Potr Krzyżanowsk 1, and Marcus Sarks 2 1 Introducton In the paper we consder a second order ellptc problem wth dscontnuous

More information

Deriving the X-Z Identity from Auxiliary Space Method

Deriving the X-Z Identity from Auxiliary Space Method Dervng the X-Z Identty from Auxlary Space Method Long Chen Department of Mathematcs, Unversty of Calforna at Irvne, Irvne, CA 92697 chenlong@math.uc.edu 1 Iteratve Methods In ths paper we dscuss teratve

More information

Non shape regular domain decompositions: an analysis using a stable decomposition in H 1 0

Non shape regular domain decompositions: an analysis using a stable decomposition in H 1 0 on shape regular doman decompostons: an analyss usng a stable decomposton n 1 0 Martn J. Gander 1, Laurence alpern, and Kévn Santugn Repquet 3 1 Unversté de Genève, Secton de Mathématques, Martn.Gander@unge.ch

More information

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix Lectures - Week 4 Matrx norms, Condtonng, Vector Spaces, Lnear Independence, Spannng sets and Bass, Null space and Range of a Matrx Matrx Norms Now we turn to assocatng a number to each matrx. We could

More information

APPENDIX A Some Linear Algebra

APPENDIX A Some Linear Algebra APPENDIX A Some Lnear Algebra The collecton of m, n matrces A.1 Matrces a 1,1,..., a 1,n A = a m,1,..., a m,n wth real elements a,j s denoted by R m,n. If n = 1 then A s called a column vector. Smlarly,

More information

Non shape regular domain decompositions: an analysis using a stable decomposition in H 1 0

Non shape regular domain decompositions: an analysis using a stable decomposition in H 1 0 Non shape regular doman decompostons: an analyss usng a stable decomposton n H 1 0 Martn Gander 1, Laurence Halpern, and Kévn Santugn Repquet 3 Abstract In ths paper, we establsh the exstence of a stable

More information

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k)

Hongyi Miao, College of Science, Nanjing Forestry University, Nanjing ,China. (Received 20 June 2013, accepted 11 March 2014) I)ϕ (k) ISSN 1749-3889 (prnt), 1749-3897 (onlne) Internatonal Journal of Nonlnear Scence Vol.17(2014) No.2,pp.188-192 Modfed Block Jacob-Davdson Method for Solvng Large Sparse Egenproblems Hongy Mao, College of

More information

2.3 Nilpotent endomorphisms

2.3 Nilpotent endomorphisms s a block dagonal matrx, wth A Mat dm U (C) In fact, we can assume that B = B 1 B k, wth B an ordered bass of U, and that A = [f U ] B, where f U : U U s the restrcton of f to U 40 23 Nlpotent endomorphsms

More information

The Finite Element Method: A Short Introduction

The Finite Element Method: A Short Introduction Te Fnte Element Metod: A Sort ntroducton Wat s FEM? Te Fnte Element Metod (FEM) ntroduced by engneers n late 50 s and 60 s s a numercal tecnque for solvng problems wc are descrbed by Ordnary Dfferental

More information

Nonlinear Overlapping Domain Decomposition Methods

Nonlinear Overlapping Domain Decomposition Methods Nonlnear Overlappng Doman Decomposton Methods Xao-Chuan Ca 1 Department of Computer Scence, Unversty of Colorado at Boulder, Boulder, CO 80309, ca@cs.colorado.edu Summary. We dscuss some overlappng doman

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Lecture 12: Discrete Laplacian

Lecture 12: Discrete Laplacian Lecture 12: Dscrete Laplacan Scrbe: Tanye Lu Our goal s to come up wth a dscrete verson of Laplacan operator for trangulated surfaces, so that we can use t n practce to solve related problems We are mostly

More information

Errors for Linear Systems

Errors for Linear Systems Errors for Lnear Systems When we solve a lnear system Ax b we often do not know A and b exactly, but have only approxmatons  and ˆb avalable. Then the best thng we can do s to solve ˆx ˆb exactly whch

More information

Relaxation Methods for Iterative Solution to Linear Systems of Equations

Relaxation Methods for Iterative Solution to Linear Systems of Equations Relaxaton Methods for Iteratve Soluton to Lnear Systems of Equatons Gerald Recktenwald Portland State Unversty Mechancal Engneerng Department gerry@pdx.edu Overvew Techncal topcs Basc Concepts Statonary

More information

TR A BDDC ALGORITHM FOR PROBLEMS WITH MORTAR DISCRETIZATION. September 4, 2005

TR A BDDC ALGORITHM FOR PROBLEMS WITH MORTAR DISCRETIZATION. September 4, 2005 TR2005-873 A BDDC ALGORITHM FOR PROBLEMS WITH MORTAR DISCRETIZATION HYEA HYUN KIM, MAKSYMILIAN DRYJA, AND OLOF B WIDLUND September 4, 2005 Abstract A BDDC balancng doman decomposton by constrants algorthm

More information

A new family of high regularity elements

A new family of high regularity elements A new famly of hgh regularty elements Jguang Sun Abstract In ths paper, we propose a new famly of hgh regularty fnte element spaces. The global approxmaton spaces are obtaned n two steps. We frst buld

More information

Spectral Graph Theory and its Applications September 16, Lecture 5

Spectral Graph Theory and its Applications September 16, Lecture 5 Spectral Graph Theory and ts Applcatons September 16, 2004 Lecturer: Danel A. Spelman Lecture 5 5.1 Introducton In ths lecture, we wll prove the followng theorem: Theorem 5.1.1. Let G be a planar graph

More information

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2 Salmon: Lectures on partal dfferental equatons 5. Classfcaton of second-order equatons There are general methods for classfyng hgher-order partal dfferental equatons. One s very general (applyng even to

More information

Robust Norm Equivalencies and Preconditioning

Robust Norm Equivalencies and Preconditioning Robust Norm Equvalences and Precondtonng Karl Scherer Insttut für Angewandte Mathematk, Unversty of Bonn, Wegelerstr. 6, 53115 Bonn, Germany Summary. In ths contrbuton we report on work done n contnuaton

More information

Numerical Heat and Mass Transfer

Numerical Heat and Mass Transfer Master degree n Mechancal Engneerng Numercal Heat and Mass Transfer 06-Fnte-Dfference Method (One-dmensonal, steady state heat conducton) Fausto Arpno f.arpno@uncas.t Introducton Why we use models and

More information

New Method for Solving Poisson Equation. on Irregular Domains

New Method for Solving Poisson Equation. on Irregular Domains Appled Mathematcal Scences Vol. 6 01 no. 8 369 380 New Method for Solvng Posson Equaton on Irregular Domans J. Izadan and N. Karamooz Department of Mathematcs Facult of Scences Mashhad BranchIslamc Azad

More information

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0 MODULE 2 Topcs: Lnear ndependence, bass and dmenson We have seen that f n a set of vectors one vector s a lnear combnaton of the remanng vectors n the set then the span of the set s unchanged f that vector

More information

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011

Stanford University CS359G: Graph Partitioning and Expanders Handout 4 Luca Trevisan January 13, 2011 Stanford Unversty CS359G: Graph Parttonng and Expanders Handout 4 Luca Trevsan January 3, 0 Lecture 4 In whch we prove the dffcult drecton of Cheeger s nequalty. As n the past lectures, consder an undrected

More information

NP-Completeness : Proofs

NP-Completeness : Proofs NP-Completeness : Proofs Proof Methods A method to show a decson problem Π NP-complete s as follows. (1) Show Π NP. (2) Choose an NP-complete problem Π. (3) Show Π Π. A method to show an optmzaton problem

More information

Appendix B. The Finite Difference Scheme

Appendix B. The Finite Difference Scheme 140 APPENDIXES Appendx B. The Fnte Dfference Scheme In ths appendx we present numercal technques whch are used to approxmate solutons of system 3.1 3.3. A comprehensve treatment of theoretcal and mplementaton

More information

Report on Image warping

Report on Image warping Report on Image warpng Xuan Ne, Dec. 20, 2004 Ths document summarzed the algorthms of our mage warpng soluton for further study, and there s a detaled descrpton about the mplementaton of these algorthms.

More information

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

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

More information

More metrics on cartesian products

More metrics on cartesian products More metrcs on cartesan products If (X, d ) are metrc spaces for 1 n, then n Secton II4 of the lecture notes we defned three metrcs on X whose underlyng topologes are the product topology The purpose of

More information

The Second Eigenvalue of Planar Graphs

The Second Eigenvalue of Planar Graphs Spectral Graph Theory Lecture 20 The Second Egenvalue of Planar Graphs Danel A. Spelman November 11, 2015 Dsclamer These notes are not necessarly an accurate representaton of what happened n class. The

More information

Finding Dense Subgraphs in G(n, 1/2)

Finding Dense Subgraphs in G(n, 1/2) Fndng Dense Subgraphs n Gn, 1/ Atsh Das Sarma 1, Amt Deshpande, and Rav Kannan 1 Georga Insttute of Technology,atsh@cc.gatech.edu Mcrosoft Research-Bangalore,amtdesh,annan@mcrosoft.com Abstract. Fndng

More information

Edge Isoperimetric Inequalities

Edge Isoperimetric Inequalities November 7, 2005 Ross M. Rchardson Edge Isopermetrc Inequaltes 1 Four Questons Recall that n the last lecture we looked at the problem of sopermetrc nequaltes n the hypercube, Q n. Our noton of boundary

More information

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space.

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space. Lnear, affne, and convex sets and hulls In the sequel, unless otherwse specfed, X wll denote a real vector space. Lnes and segments. Gven two ponts x, y X, we defne xy = {x + t(y x) : t R} = {(1 t)x +

More information

Affine transformations and convexity

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

More information

partial dierential equations by iteratively solving subproblems dened on and localized treatment of complex and irregular geometries, singularities

partial dierential equations by iteratively solving subproblems dened on and localized treatment of complex and irregular geometries, singularities Acta Numerca (1994), pp. 61{143 Doman decomposton algorthms Tony F. Chan Department of Mathematcs, Unversty of Calforna at Los Angeles, Los Angeles, CA 90024, USA Emal: chan@math.ucla.edu. Tarek P. Mathew

More information

Convexity preserving interpolation by splines of arbitrary degree

Convexity preserving interpolation by splines of arbitrary degree Computer Scence Journal of Moldova, vol.18, no.1(52), 2010 Convexty preservng nterpolaton by splnes of arbtrary degree Igor Verlan Abstract In the present paper an algorthm of C 2 nterpolaton of dscrete

More information

Vector Norms. Chapter 7 Iterative Techniques in Matrix Algebra. Cauchy-Bunyakovsky-Schwarz Inequality for Sums. Distances. Convergence.

Vector Norms. Chapter 7 Iterative Techniques in Matrix Algebra. Cauchy-Bunyakovsky-Schwarz Inequality for Sums. Distances. Convergence. Vector Norms Chapter 7 Iteratve Technques n Matrx Algebra Per-Olof Persson persson@berkeley.edu Department of Mathematcs Unversty of Calforna, Berkeley Math 128B Numercal Analyss Defnton A vector norm

More information

5 The Rational Canonical Form

5 The Rational Canonical Form 5 The Ratonal Canoncal Form Here p s a monc rreducble factor of the mnmum polynomal m T and s not necessarly of degree one Let F p denote the feld constructed earler n the course, consstng of all matrces

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

More information

Lecture 3. Ax x i a i. i i

Lecture 3. Ax x i a i. i i 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

Problem Set 9 Solutions

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

More information

CSCE 790S Background Results

CSCE 790S Background Results CSCE 790S Background Results Stephen A. Fenner September 8, 011 Abstract These results are background to the course CSCE 790S/CSCE 790B, Quantum Computaton and Informaton (Sprng 007 and Fall 011). Each

More information

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b

A New Refinement of Jacobi Method for Solution of Linear System Equations AX=b Int J Contemp Math Scences, Vol 3, 28, no 17, 819-827 A New Refnement of Jacob Method for Soluton of Lnear System Equatons AX=b F Naem Dafchah Department of Mathematcs, Faculty of Scences Unversty of Gulan,

More information

Case A. P k = Ni ( 2L i k 1 ) + (# big cells) 10d 2 P k.

Case A. P k = Ni ( 2L i k 1 ) + (# big cells) 10d 2 P k. THE CELLULAR METHOD In ths lecture, we ntroduce the cellular method as an approach to ncdence geometry theorems lke the Szemeréd-Trotter theorem. The method was ntroduced n the paper Combnatoral complexty

More information

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 )

Yong Joon Ryang. 1. Introduction Consider the multicommodity transportation problem with convex quadratic cost function. 1 2 (x x0 ) T Q(x x 0 ) Kangweon-Kyungk Math. Jour. 4 1996), No. 1, pp. 7 16 AN ITERATIVE ROW-ACTION METHOD FOR MULTICOMMODITY TRANSPORTATION PROBLEMS Yong Joon Ryang Abstract. The optmzaton problems wth quadratc constrants often

More information

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS)

Some Comments on Accelerating Convergence of Iterative Sequences Using Direct Inversion of the Iterative Subspace (DIIS) Some Comments on Acceleratng Convergence of Iteratve Sequences Usng Drect Inverson of the Iteratve Subspace (DIIS) C. Davd Sherrll School of Chemstry and Bochemstry Georga Insttute of Technology May 1998

More information

8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS

8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 493 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces you have studed thus far n the text are real vector spaces because the scalars

More information

Polynomial Regression Models

Polynomial Regression Models LINEAR REGRESSION ANALYSIS MODULE XII Lecture - 6 Polynomal Regresson Models Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur Test of sgnfcance To test the sgnfcance

More information

Robust FETI solvers for multiscale elliptic PDEs

Robust FETI solvers for multiscale elliptic PDEs Robust FETI solvers for multscale ellptc PDEs Clemens Pechsten 1 and Robert Schechl 2 Abstract Fnte element tearng and nterconnectng (FETI) methods are effcent parallel doman decomposton solvers for large-scale

More information

Bath Institute For Complex Systems

Bath Institute For Complex Systems BICS Bath Insttute for Complex Systems Analyss of FETI methods for multscale PDEs Part II: Interface varaton C. Pechsten and R. Schechl Bath Insttute For Complex Systems Preprnt 7/09 (2009) http://www.bath.ac.uk/math-sc/bics

More information

Maximizing the number of nonnegative subsets

Maximizing the number of nonnegative subsets Maxmzng the number of nonnegatve subsets Noga Alon Hao Huang December 1, 213 Abstract Gven a set of n real numbers, f the sum of elements of every subset of sze larger than k s negatve, what s the maxmum

More information

General viscosity iterative method for a sequence of quasi-nonexpansive mappings

General viscosity iterative method for a sequence of quasi-nonexpansive mappings Avalable onlne at www.tjnsa.com J. Nonlnear Sc. Appl. 9 (2016), 5672 5682 Research Artcle General vscosty teratve method for a sequence of quas-nonexpansve mappngs Cuje Zhang, Ynan Wang College of Scence,

More information

STAT 309: MATHEMATICAL COMPUTATIONS I FALL 2018 LECTURE 16

STAT 309: MATHEMATICAL COMPUTATIONS I FALL 2018 LECTURE 16 STAT 39: MATHEMATICAL COMPUTATIONS I FALL 218 LECTURE 16 1 why teratve methods f we have a lnear system Ax = b where A s very, very large but s ether sparse or structured (eg, banded, Toepltz, banded plus

More information

Downloaded 07/31/13 to Redistribution subject to SIAM license or copyright; see

Downloaded 07/31/13 to Redistribution subject to SIAM license or copyright; see SIAM J NUMER ANAL Vol 47, No 1, pp 136 157 c 2008 Socety for Industral and Appled Mathematcs A BDDC METHOD FOR MORTAR DISCRETIZATIONS USING A TRANSFORMATION OF BASIS HYEA HYUN KIM, MAKSYMILIAN DRYJA, AND

More information

MAT 578 Functional Analysis

MAT 578 Functional Analysis MAT 578 Functonal Analyss John Qugg Fall 2008 Locally convex spaces revsed September 6, 2008 Ths secton establshes the fundamental propertes of locally convex spaces. Acknowledgment: although I wrote these

More information

TR A FETI-DP FORMULATION OF THREE DIMENSIONAL ELASTICITY PROBLEMS WITH MORTAR DISCRETIZATION. April 26, 2005

TR A FETI-DP FORMULATION OF THREE DIMENSIONAL ELASTICITY PROBLEMS WITH MORTAR DISCRETIZATION. April 26, 2005 TR2005-863 A FETI-DP FORMULATION OF THREE DIMENSIONAL ELASTICITY PROBLEMS WITH MORTAR DISCRETIZATION HYEA HYUN KIM Aprl 26, 2005 Abstract. In ths paper, a FETI-DP formulaton for the three dmensonal elastcty

More information

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal Inner Product Defnton 1 () A Eucldean space s a fnte-dmensonal vector space over the reals R, wth an nner product,. Defnton 2 (Inner Product) An nner product, on a real vector space X s a symmetrc, blnear,

More information

Bézier curves. Michael S. Floater. September 10, These notes provide an introduction to Bézier curves. i=0

Bézier curves. Michael S. Floater. September 10, These notes provide an introduction to Bézier curves. i=0 Bézer curves Mchael S. Floater September 1, 215 These notes provde an ntroducton to Bézer curves. 1 Bernsten polynomals Recall that a real polynomal of a real varable x R, wth degree n, s a functon of

More information

= = = (a) Use the MATLAB command rref to solve the system. (b) Let A be the coefficient matrix and B be the right-hand side of the system.

= = = (a) Use the MATLAB command rref to solve the system. (b) Let A be the coefficient matrix and B be the right-hand side of the system. Chapter Matlab Exercses Chapter Matlab Exercses. Consder the lnear system of Example n Secton.. x x x y z y y z (a) Use the MATLAB command rref to solve the system. (b) Let A be the coeffcent matrx and

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

Isogeometric Analysis with Geometrically Continuous Functions on Multi-Patch Geometries. Mario Kapl, Florian Buchegger, Michel Bercovier, Bert Jüttler

Isogeometric Analysis with Geometrically Continuous Functions on Multi-Patch Geometries. Mario Kapl, Florian Buchegger, Michel Bercovier, Bert Jüttler Isogeometrc Analyss wth Geometrcally Contnuous Functons on Mult-Patch Geometres Maro Kapl Floran Buchegger Mchel Bercover Bert Jüttler G+S Report No 35 August 05 Isogeometrc Analyss wth Geometrcally Contnuous

More information

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

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

More information

Inexact Newton Methods for Inverse Eigenvalue Problems

Inexact Newton Methods for Inverse Eigenvalue Problems Inexact Newton Methods for Inverse Egenvalue Problems Zheng-jan Ba Abstract In ths paper, we survey some of the latest development n usng nexact Newton-lke methods for solvng nverse egenvalue problems.

More information

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016

U.C. Berkeley CS294: Spectral Methods and Expanders Handout 8 Luca Trevisan February 17, 2016 U.C. Berkeley CS94: Spectral Methods and Expanders Handout 8 Luca Trevsan February 7, 06 Lecture 8: Spectral Algorthms Wrap-up In whch we talk about even more generalzatons of Cheeger s nequaltes, and

More information

Exercise Solutions to Real Analysis

Exercise Solutions to Real Analysis xercse Solutons to Real Analyss Note: References refer to H. L. Royden, Real Analyss xersze 1. Gven any set A any ɛ > 0, there s an open set O such that A O m O m A + ɛ. Soluton 1. If m A =, then there

More information

Every planar graph is 4-colourable a proof without computer

Every planar graph is 4-colourable a proof without computer Peter Dörre Department of Informatcs and Natural Scences Fachhochschule Südwestfalen (Unversty of Appled Scences) Frauenstuhlweg 31, D-58644 Iserlohn, Germany Emal: doerre(at)fh-swf.de Mathematcs Subject

More information

Perron Vectors of an Irreducible Nonnegative Interval Matrix

Perron Vectors of an Irreducible Nonnegative Interval Matrix Perron Vectors of an Irreducble Nonnegatve Interval Matrx Jr Rohn August 4 2005 Abstract As s well known an rreducble nonnegatve matrx possesses a unquely determned Perron vector. As the man result of

More information

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

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

More information

The L(2, 1)-Labeling on -Product of Graphs

The L(2, 1)-Labeling on -Product of Graphs Annals of Pure and Appled Mathematcs Vol 0, No, 05, 9-39 ISSN: 79-087X (P, 79-0888(onlne Publshed on 7 Aprl 05 wwwresearchmathscorg Annals of The L(, -Labelng on -Product of Graphs P Pradhan and Kamesh

More information

MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS

MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS MATH 241B FUNCTIONAL ANALYSIS - NOTES EXAMPLES OF C ALGEBRAS These are nformal notes whch cover some of the materal whch s not n the course book. The man purpose s to gve a number of nontrval examples

More information

CME 302: NUMERICAL LINEAR ALGEBRA FALL 2005/06 LECTURE 13

CME 302: NUMERICAL LINEAR ALGEBRA FALL 2005/06 LECTURE 13 CME 30: NUMERICAL LINEAR ALGEBRA FALL 005/06 LECTURE 13 GENE H GOLUB 1 Iteratve Methods Very large problems (naturally sparse, from applcatons): teratve methods Structured matrces (even sometmes dense,

More information

Lecture Notes on Linear Regression

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

More information

MMA and GCMMA two methods for nonlinear optimization

MMA and GCMMA two methods for nonlinear optimization MMA and GCMMA two methods for nonlnear optmzaton Krster Svanberg Optmzaton and Systems Theory, KTH, Stockholm, Sweden. krlle@math.kth.se Ths note descrbes the algorthms used n the author s 2007 mplementatons

More information

The interface control domain decomposition (ICDD) method for the Stokes problem. (Received: 15 July Accepted: 13 September 2013)

The interface control domain decomposition (ICDD) method for the Stokes problem. (Received: 15 July Accepted: 13 September 2013) Journal of Coupled Systems Multscale Dynamcs Copyrght 2013 by Amercan Scentfc Publshers All rghts reserved. Prnted n the Unted States of Amerca do:10.1166/jcsmd.2013.1026 J. Coupled Syst. Multscale Dyn.

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

Kernel Methods and SVMs Extension

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

More information

The Geometry of Logit and Probit

The Geometry of Logit and Probit The Geometry of Logt and Probt Ths short note s meant as a supplement to Chapters and 3 of Spatal Models of Parlamentary Votng and the notaton and reference to fgures n the text below s to those two chapters.

More information

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography

CSci 6974 and ECSE 6966 Math. Tech. for Vision, Graphics and Robotics Lecture 21, April 17, 2006 Estimating A Plane Homography CSc 6974 and ECSE 6966 Math. Tech. for Vson, Graphcs and Robotcs Lecture 21, Aprl 17, 2006 Estmatng A Plane Homography Overvew We contnue wth a dscusson of the major ssues, usng estmaton of plane projectve

More information

Some basic inequalities. Definition. Let V be a vector space over the complex numbers. An inner product is given by a function, V V C

Some basic inequalities. Definition. Let V be a vector space over the complex numbers. An inner product is given by a function, V V C Some basc nequaltes Defnton. Let V be a vector space over the complex numbers. An nner product s gven by a functon, V V C (x, y) x, y satsfyng the followng propertes (for all x V, y V and c C) (1) x +

More information

EFFICIENT DOMAIN DECOMPOSITION METHOD FOR ACOUSTIC SCATTERING IN MULTI-LAYERED MEDIA

EFFICIENT DOMAIN DECOMPOSITION METHOD FOR ACOUSTIC SCATTERING IN MULTI-LAYERED MEDIA European Conference on Computatonal Flud Dynamcs ECCOMAS CFD 2006 P. Wesselng, E. Oñate and J. Péraux (Eds) c TU Delft, The Netherlands, 2006 EFFICIENT DOMAIN DECOMPOSITION METHOD FOR ACOUSTIC SCATTERING

More information

Asymptotics of the Solution of a Boundary Value. Problem for One-Characteristic Differential. Equation Degenerating into a Parabolic Equation

Asymptotics of the Solution of a Boundary Value. Problem for One-Characteristic Differential. Equation Degenerating into a Parabolic Equation Nonl. Analyss and Dfferental Equatons, ol., 4, no., 5 - HIKARI Ltd, www.m-har.com http://dx.do.org/.988/nade.4.456 Asymptotcs of the Soluton of a Boundary alue Problem for One-Characterstc Dfferental Equaton

More information

Opus: University of Bath Online Publication Store

Opus: University of Bath Online Publication Store Pechsten, C. and Schechl, R. (0) Analyss of FETI methods for multscale PDEs. Part II: nterface varaton. Numersche Mathematk, 8 (3). pp. 485-59. ISSN 009-599X Lnk to offcal URL (f avalable): http://dx.do.org/0.007/s00-0-0359-

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

The lower and upper bounds on Perron root of nonnegative irreducible matrices

The lower and upper bounds on Perron root of nonnegative irreducible matrices Journal of Computatonal Appled Mathematcs 217 (2008) 259 267 wwwelsevercom/locate/cam The lower upper bounds on Perron root of nonnegatve rreducble matrces Guang-Xn Huang a,, Feng Yn b,keguo a a College

More information

A new Approach for Solving Linear Ordinary Differential Equations

A new Approach for Solving Linear Ordinary Differential Equations , ISSN 974-57X (Onlne), ISSN 974-5718 (Prnt), Vol. ; Issue No. 1; Year 14, Copyrght 13-14 by CESER PUBLICATIONS A new Approach for Solvng Lnear Ordnary Dfferental Equatons Fawz Abdelwahd Department of

More information

FACTORIZATION IN KRULL MONOIDS WITH INFINITE CLASS GROUP

FACTORIZATION IN KRULL MONOIDS WITH INFINITE CLASS GROUP C O L L O Q U I U M M A T H E M A T I C U M VOL. 80 1999 NO. 1 FACTORIZATION IN KRULL MONOIDS WITH INFINITE CLASS GROUP BY FLORIAN K A I N R A T H (GRAZ) Abstract. Let H be a Krull monod wth nfnte class

More information

NUMERICAL DIFFERENTIATION

NUMERICAL DIFFERENTIATION NUMERICAL DIFFERENTIATION 1 Introducton Dfferentaton s a method to compute the rate at whch a dependent output y changes wth respect to the change n the ndependent nput x. Ths rate of change s called the

More information

Complete subgraphs in multipartite graphs

Complete subgraphs in multipartite graphs Complete subgraphs n multpartte graphs FLORIAN PFENDER Unverstät Rostock, Insttut für Mathematk D-18057 Rostock, Germany Floran.Pfender@un-rostock.de Abstract Turán s Theorem states that every graph G

More information

Bezier curves. Michael S. Floater. August 25, These notes provide an introduction to Bezier curves. i=0

Bezier curves. Michael S. Floater. August 25, These notes provide an introduction to Bezier curves. i=0 Bezer curves Mchael S. Floater August 25, 211 These notes provde an ntroducton to Bezer curves. 1 Bernsten polynomals Recall that a real polynomal of a real varable x R, wth degree n, s a functon of the

More information

THE WEIGHTED WEAK TYPE INEQUALITY FOR THE STRONG MAXIMAL FUNCTION

THE WEIGHTED WEAK TYPE INEQUALITY FOR THE STRONG MAXIMAL FUNCTION THE WEIGHTED WEAK TYPE INEQUALITY FO THE STONG MAXIMAL FUNCTION THEMIS MITSIS Abstract. We prove the natural Fefferman-Sten weak type nequalty for the strong maxmal functon n the plane, under the assumpton

More information

Another converse of Jensen s inequality

Another converse of Jensen s inequality Another converse of Jensen s nequalty Slavko Smc Abstract. We gve the best possble global bounds for a form of dscrete Jensen s nequalty. By some examples ts frutfulness s shown. 1. Introducton Throughout

More information

Vapnik-Chervonenkis theory

Vapnik-Chervonenkis theory Vapnk-Chervonenks theory Rs Kondor June 13, 2008 For the purposes of ths lecture, we restrct ourselves to the bnary supervsed batch learnng settng. We assume that we have an nput space X, and an unknown

More information

MATH 5630: Discrete Time-Space Model Hung Phan, UMass Lowell March 1, 2018

MATH 5630: Discrete Time-Space Model Hung Phan, UMass Lowell March 1, 2018 MATH 5630: Dscrete Tme-Space Model Hung Phan, UMass Lowell March, 08 Newton s Law of Coolng Consder the coolng of a well strred coffee so that the temperature does not depend on space Newton s law of collng

More information

Lecture 10 Support Vector Machines II

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

More information

Math 217 Fall 2013 Homework 2 Solutions

Math 217 Fall 2013 Homework 2 Solutions Math 17 Fall 013 Homework Solutons Due Thursday Sept. 6, 013 5pm Ths homework conssts of 6 problems of 5 ponts each. The total s 30. You need to fully justfy your answer prove that your functon ndeed has

More information

DIFFERENTIAL FORMS BRIAN OSSERMAN

DIFFERENTIAL FORMS BRIAN OSSERMAN DIFFERENTIAL FORMS BRIAN OSSERMAN Dfferentals are an mportant topc n algebrac geometry, allowng the use of some classcal geometrc arguments n the context of varetes over any feld. We wll use them to defne

More information

SUCCESSIVE MINIMA AND LATTICE POINTS (AFTER HENK, GILLET AND SOULÉ) M(B) := # ( B Z N)

SUCCESSIVE MINIMA AND LATTICE POINTS (AFTER HENK, GILLET AND SOULÉ) M(B) := # ( B Z N) SUCCESSIVE MINIMA AND LATTICE POINTS (AFTER HENK, GILLET AND SOULÉ) S.BOUCKSOM Abstract. The goal of ths note s to present a remarably smple proof, due to Hen, of a result prevously obtaned by Gllet-Soulé,

More information

On a direct solver for linear least squares problems

On a direct solver for linear least squares problems ISSN 2066-6594 Ann. Acad. Rom. Sc. Ser. Math. Appl. Vol. 8, No. 2/2016 On a drect solver for lnear least squares problems Constantn Popa Abstract The Null Space (NS) algorthm s a drect solver for lnear

More information

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES

VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES VARIATION OF CONSTANT SUM CONSTRAINT FOR INTEGER MODEL WITH NON UNIFORM VARIABLES BÂRZĂ, Slvu Faculty of Mathematcs-Informatcs Spru Haret Unversty barza_slvu@yahoo.com Abstract Ths paper wants to contnue

More information

2.29 Numerical Fluid Mechanics

2.29 Numerical Fluid Mechanics REVIEW Lecture 10: Sprng 2015 Lecture 11 Classfcaton of Partal Dfferental Equatons PDEs) and eamples wth fnte dfference dscretzatons Parabolc PDEs Ellptc PDEs Hyperbolc PDEs Error Types and Dscretzaton

More information

One-sided finite-difference approximations suitable for use with Richardson extrapolation

One-sided finite-difference approximations suitable for use with Richardson extrapolation Journal of Computatonal Physcs 219 (2006) 13 20 Short note One-sded fnte-dfference approxmatons sutable for use wth Rchardson extrapolaton Kumar Rahul, S.N. Bhattacharyya * Department of Mechancal Engneerng,

More information