A Recursive Blocked Schur Algorithm for Computing the Matrix Square Root. Deadman, Edvin and Higham, Nicholas J. and Ralha, Rui. MIMS EPrint: 2012.
|
|
- Gwenda Hubbard
- 5 years ago
- Views:
Transcription
1 A Recursive Blocked Schur Algorithm for Computig the Matrix Square Root Deadma, Edvi ad Higham, Nicholas J. ad Ralha, Rui 212 MIMS EPrit: Machester Istitute for Mathematical Scieces School of Mathematics The Uiversity of Machester Reports available from: Ad by cotactig: The MIMS Secretary School of Mathematics The Uiversity of Machester Machester, M13 9PL, UK ISSN
2 A Recursive Blocked Schur Algorithm for Computig the Matrix Square Root Edvi Deadma Uiversity of Machester Numerical Algorithms Group Nicholas J. Higham Uiversity of Machester Rui Ralha Uiversity of Miho, Portugal Abstract The Schur method for computig a matrix square root reduces the matrix to the Schur triagular form ad the computes a square root of the triagular matrix. We show that by usig a recursive blockig techique the computatio of the square root of the triagular matrix ca be made rich i matrix multiplicatio. Numerical experimets makig appropriate use of level 3 BLAS show sigificat speedups over the poit algorithm, both i the square root phase ad i the algorithm as a whole. The excellet umerical stability of the poit algorithm is show to be preserved by recursive blockig. These results are exteded to the real Schur method. Recursive blockig is also show to be effective for multiplyig triagular matrices. I. INTRODUCTION A square root of matrix A C is ay matrix satisfyig X 2 A. Matrix square roots have may applicatios, icludig i Markov models of fiace, the solutio of differetial equatios ad the computatio of the polar decompositio ad the matrix sig fuctio [9]. A square root of a matrix (if oe exists) is ot uique. However, if A has o eigevalues o the closed egative real lie the there is a uique pricipal square root A 1/2 whose eigevalues all lie i the ope right half-plae. If A is real, the so is A 1/2. For proofs of these facts ad more o the theory of matrix square roots see [9]. The most umerically stable way of computig matrix square roots is via the Schur method of Björck ad Hammarlig [3]. The matrix A is reduced to upper triagular form ad a recurrece relatio eables the square root of the triagular matrix to be computed a colum or superdiagoal at a time. I II we show that the recurrece ca be reorgaized usig a stadard blockig scheme or recursive blockig i order to make it rich i matrix multiplicatios. We show experimetally that sigificat speedups result whe level 3 BLAS are exploited i the implemetatio. I III we show that the blocked methods maitai the excellet backward stability of the o-blocked method. I IV we discuss the use of the ew approach withi the Schur method ad explai how it ca be exteded to the real Schur method of Higham [7]. We compare our implemetatios writte for the NAG Library with existig MATLAB fuctios. Fially, i V we discuss some further applicatios of recursive blockig for triagular matrices. II. THE USE OF BLOCKING IN THE SCHUR METHOD To compute A 1/2, a Schur decompositio A QT Q is obtaied, where T is upper triagular ad Q is uitary. The A 1/2 QT 1/2 Q. For the remaider of this sectio we will focus o upper triagular matrices oly. The equatio U 2 T (1) ca be solved by otig that U is also upper triagular, so that by equatig elemets, U ii U ij + U ij U jj T ij U 2 ii T ii, (2) j 1 ki+1 U ik U kj. (3) These equatios ca be solved either a colum or a superdiagoal at a time. Differet choices of sig i the scalar square roots of (2) lead to differet matrix square roots. This method will be referred to hereafter as the poit method. The algorithm ca be blocked by lettig the U ij ad T ij i (2) ad (3) refer to m m blocks, where m. The diagoal blocks U ii are the obtaied usig the poit method ad the off-diagoal blocks are obtaied by solvig the Sylvester equatios (3) usig LAPACK routie xtrsyl (where x deotes D or Z accordig to whether real or complex arithmetic is used) [2]. Level 3 BLAS ca be used i computig the righthad side of (3) so sigificat improvemets i efficiecy are expected. This approach is referred to as the block method. To test this approach, a Fortra implemetatio was writte ad compiled with gfortra o a 64 bit Itel i3 dual-core machie, usig the ACML Library for LAPACK ad BLAS calls. Complex upper triagular matrices were geerated, with radom elemets whose real ad imagiary parts were chose from the uiform distributio o [, 1). Figure 1 shows the ru times for the methods, for various differet sized matrices. A block size of 32 was chose, although the speed did ot appear to be particularly sesitive to the block size similar results were obtaied with blocks of size 16, 64 ad 128. The block method was foud to be up to 17 times faster tha the poit method. The residuals Û 2 T / T, where Û is the computed value of U, were similar for both methods. Table II shows that, for 496, over 9% of the ru time is spet i ZGEMM calls.
3 6 TABLE I Profilig of the block method for computig the square root of a triagular matrix, with 496. Format: time i secods (umber of calls). 5 4 Total time take: 3.74 Calls to loopig method:.9 (128) Calls to ZTRSYL 2.4 (8128) Calls to ZGEMM with 32: (341376) 3 poit block recursio TABLE II Profilig of the recursive method for computig the square root of a triagular matrix, with 496. Format: time i secods (umber of calls) Fig. 1. Ru times for the poit, block ad recursio methods for computig the square root of a complex triagular matrix. Total time take: 27.4 Calls to loopig method:.6 (128) Calls to ZTRSYL 2.22 (8128) Calls to ZGEMM total: 22.6 (1668) Calls to ZGEMM with (4) Calls to ZGEMM with (24) Calls to ZGEMM with (112) Calls to ZGEMM with (48) Calls to ZGEMM with (1984) Calls to ZGEMM with (864) A larger block size eables larger GEMM calls to be made. However, it leads to larger calls to the poit algorithm ad to xtrsyl (which oly uses level 2 BLAS). A recursive approach may allow icreased use of level 3 BLAS. Equatio (1) ca be rewritte as ( ) 2 ( ) U11 U 12 T11 T 12, (4) U 22 T 22 where the submatrices are of size /2 or (±1)/2 depedig o the parity of. The U11 2 T 11 ad U22 2 T 22 ca be solved recursively, util some base level is reached, at which poit the poit algorithm is used. The Sylvester equatio U 11 U 12 + U 12 U 22 T 12 ca the be solved usig a recursive algorithm devised by Josso ad Kågström [11]. I this algorithm, the Sylvester equatio AX + XB C, with A ad B triagular, is writte as ( ) ( ) A11 A 12 X11 X 12 + A 22 X 21 X 22 ( ) ( ) X11 X 12 B11 B 12 X 21 X 22 B 22 ( ) C11 C 12, C 21 C 22 where each submatrix is of size /2 or ( ± 1)/2. The A 11 X 11 + X 11 B 11 C 11 A 12 X 21, (5) A 11 X 12 + X 12 B 22 C 12 A 12 X 22 X 11 B 12, (6) A 22 X 21 + X 21 B 11 C 21, (7) A 22 X 22 + X 22 B 22 C 22 X 21 B 12. (8) Equatio (7) is solved recursively, followed by (5) ad (8), ad fially (6). At the base level a routie such as xtrsyl is used. The ru times for a Fortra implemetatio of the recursio method i complex arithmetic, with a base level of size 32, are show i Figure 1. The approach was foud to be faster tha the block method, ad up to 19 times faster tha the poit method, with similar residuals i each case. The precise choice of base level made little differece to the ru time. Table II shows that the ru time is domiated by GEMM calls ad that the time spet i ZTRSYL ad the poit algorithm is similar to the block method. The largest GEMM call uses a submatrix of size /4. III. STABILITY OF THE BLOCKED ALGORITHMS We use the stadard model of floatig poit arithmetic [8, 2.2] i which the result of a floatig poit operatio, op, o two scalars x ad y is writte as fl(x op y) (x op y)(1 + δ), δ u, where u is the uit roudoff. I aalyzig a sequece of floatig poit operatios it is useful to write [8, 3.4] (1 + δ i ) ρi 1 + θ, ρ i ±1, where i1 θ u 1 u : γ. It is also coveiet to defie γ γ c for some small iteger c whose precise value is uimportat. We use a hat deote a computed quatity ad write A for the matrix whose elemets are the absolute values of the elemets of A. Björck ad Hammarlig [3] obtaied a ormwise backward error boud for the Schur method. The computed square root X of the full matrix A satisfies X 2 A + A, where A F γ 3 X 2 F. (9) Higham [9, 6.2] obtaied a compoetwise boud for the triagular phase of the algorithm. The computed square root Û of the triagular matrix T satisfies Û 2 T + T, where T γ Û 2. (1)
4 This boud implies (9). We ow ivestigate whether the boud (1) still holds whe the triagular phase of the algorithm is blocked. Cosider the Sylvester equatio AX + XB C i matrices with triagular A ad B. Whe it is solved i the stadard way by the solutio of triagular systems the residual of the computed ˆX satisfies [8, 16.1] C (A X + XB) γ ( A X + X B ). (11) I the (o-recursive) block method, to boud T ij we must accout for the error i performig the matrix multiplicatios o the right-had side of (3). Stadard error aalysis for matrix multiplicatio yields, for blocks of size m, ( j 1 ) fl j 1 Û ik Û kj Û ik Û kj γ Û 2 ij. ki+1 ki+1 Substitutig this ito the residual for the Sylvester equatio i the off-diagoal blocks, we obtai the compoetwise boud (1). To obtai a boud for the recursive blocked method we must first check if (11) holds whe the Sylvester equatio is solved usig Josso ad Kågström s recursive algorithm. This ca be doe by iductio, assumig that (11) holds at the base level. For the iductive step, if suffices to icorporate the error estimates for the matrix multiplicatios i the right had sides of (5) (8) ito the residual boud. Iductio ca the be applied to the recursive blocked method for the square root. The bouds (1) ad (11) are assumed to hold at the base level. The iductive step is similar to the aalysis for the block method. Overall, (1) is obtaied. We coclude that both our blocked algorithms for computig the matrix square root satisfy backward error bouds of the same forms (9) ad (1) as the poit algorithm. IV. IMPLEMENTATION Whe used with full (o-triagular) matrices, more modest speedups are expected because of the sigificat overhead i computig the Schur decompositio. Figure 2 compares ru times of the MATLAB fuctio sqrtm (which does ot use ay blockig) with Fortra implemetatios of the recursive blocked method (fort_recurse) ad the poit algorithm (fort_poit), called from withi MATLAB usig a mex iterface. The recursive routie is foud to be up to 2.5 times faster tha sqrtm ad 2 times faster tha fort_poit. A extesio of the Schur method due to Higham [7] eables the square root of a real matrix to be computed without usig complex arithmetic. A real Schur decompositio of A is computed. Square roots of the 2 2 diagoal blocks of the upper quasi-triagular factor are computed usig a explicit formula. The recursio (3) ow proceeds either a block colum or a block superdiagoal at a time, where the blocks are of size 1 1, 1 2, 2 1 or 2 2 depedig o the diagoal block structure. A MATLAB implemetatio of this algorithm sqrtm_real is available i the Matrix Fuctio Toolbox [6]. The algorithm ca also be implemeted i a recursive sqrtm fort_poit fort_recurse Fig. 2. Ru times for sqrtm, fort_recurse ad fort_poit for computig the square root of a full matrix with elemets whose real ad imagiary parts are chose from the uiform radom distributio o the iterval [, 1). maer, the oly subtlety beig that the splittig poit for the recursio must be chose to avoid splittig ay 2 2 diagoal blocks. A similar error aalysis to III applies to the real recursive method, though sice oly a ormwise boud is available for the poit algorithm applied to the quasi-triagular matrix the backward error boud (1) holds i the Frobeius orm rather tha elemetwise. Figure 3 compares the ru times of sqrtm ad sqrtm_real with Fortra implemetatios of the real recursive method (fort_recurse_real) ad the real poit method (fort_poit_real), also called from withi MAT- LAB. The recursive routie is foud to be up to 6 times faster tha sqrtm ad sqrtm_real ad 2 times faster tha fort_poit_real. Both the real ad complex recursive blocked routies sped over 9% of their ru time i computig the Schur decompositio, compared with 25% for sqrtm, 44% for fort_poit, 16% for sqrtm_real ad 46% for fort_poit_real. The latter four percetages reflect the overhead of the MAT- LAB iterpreter i executig the recurreces for the (quasi-) triagular square root phase. V. FURTHER APPLICATIONS OF RECURSIVE BLOCKING We briefly metio two further applicatios of recursive blockig schemes. Curretly there are o LAPACK or BLAS routies desiged specifically for multiplyig two triagular matrices, T U V (the closest is the BLAS routie xtrmm which multiplies a triagular matrix by a full matrix). However, a block algorithm is easily derived by partitioig the matrices ito blocks. The product of two off-diagoal blocks is computed usig xgemm. The product of a off-diagoal block ad a diagoal block is computed usig xtrmm. Fially the poit method is used whe multiplyig two diagoal blocks.
5 sqrtm_real sqrtm fort_poit_real fort_recurse_real The iverse of a triagular matrix ca be computed recursively, by expadig UU 1 I as ( ) ( ) ( ) U11 U 12 (U 1 ) 11 (U 1 ) 12 I U 22 (U 1. ) 22 I The (Û 1 ) 11 ad (Û 1 ) 22 are computed recursively ad (Û 1 ) 12 is obtaied by solvig U 11 (Û 1 ) 12 + U 12 (Û 1 ) 22. Provided that forward substitutio is used, the right (or left) recursive iversio method ca be show iductively to satisfy the same right (or left) elemetwise residual boud as the poit method [4]. A Fortra implemetatio of this idea was foud to perform similarly to LAPACK code xtrtri, so o real beefit was derived from recursive blockig Fig. 3. Ru times for sqrtm, sqrtm_real, fort_recurse_real ad fort_poit_real for computig the square root of a full matrix with elemets chose from the uiform radom distributio o [, 1) poit block recursio Fig. 4. Ru times for the poit, block ad recursio methods for multiplyig radomly geerated triagular matrices. I the recursive approach, T UV is rewritte as ( ) ( ) ( ) T11 T 12 U11 U 12 V11 V 12. T 22 U 22 V 22 The T 11 U 11 V 11 ad T 22 U 22 V 22 are computed recursively ad T 12 U 11 V 12 + U 12 V 22 is computed usig two calls to xtrmm. Figure 4 shows ru times for some triagular matrix multiplicatios usig Fortra implemetatios of the poit method, stadard blockig ad recursive blockig (the block size ad base levels were both 32 i this case, although the results were ot too sesitive to the precise choice of these parameters). As for the matrix square root, the block algorithms sigificatly outperform the poit algorithm, with the recursive approach performig the best. VI. CONCLUSIONS We ivestigated two differet blockig techiques withi Björck ad Hammarlig s recurrece for computig a square root of a triagular matrix, fidig that recursive blockig gives the best performace. Neither approach etails ay loss of backward stability. We implemeted the recursive blockig with both the Schur method ad the real Schur method (which works etirely i real arithmetic) ad foud the ew codes to be sigificatly faster tha correspodig poit codes, which iclude the MATLAB fuctios sqrtm (built-i) ad sqrtm_real (from [6]). The ew codes will appear i the ext mark of the NAG Library [12]. Recursive blockig is also fruitful for multiplyig triagular matrices. Because of the importace of the (quasi-) triagular square root, which arises i algorithms for computig the matrix logarithm [1], matrix pth roots [5], ad arbitrary matrix powers [1], this computatioal kerel is a strog coteder for iclusio i ay future extesios of the BLAS. REFERENCES [1] A. H. Al-Mohy ad N. J. Higham. Improved iverse scalig ad squarig algorithms for the matrix logarithm. MIMS EPrit , The Uiversity of Machester, Oct [2] E. Aderso, Z. Bai, C. Bischof, S. Blackford, J. Demmel, J. Dogarra, J. Du Croz, A. Greebaum, S. Hammarlig, A. McKeey, ad D. Sorese. LAPACK Users Guide. Society for Idustrial ad Applied Mathematics, Philadelphia, PA, third editio, [3] Å. Björck ad S. Hammarlig. A Schur method for the square root of a matrix. Liear Algebra Appl., 52/53:127 14, [4] J. J. Du Croz ad N. J. Higham. Stability methods for matrix iversio. IMA J. Numer. Aal., 12(1):1 19, [5] C.-H. Guo ad N. J. Higham. A Schur Newto method for the matrix pth root ad its iverse. SIAM J. Matrix Aal. Appl., 28(3):788 84, 6. [6] N. J. Higham. The Matrix Fuctio Toolbox. ac.uk/ higham/mftoolbox. [7] N. J. Higham. Computig real square roots of a real matrix. Liear Algebra Appl., 88/89:45 43, [8] N. J. Higham. Accuracy ad Stability of Numerical Algorithms. SIAM, 2d editio, 2. [9] N. J. Higham. Fuctios of Matrices: Theory ad Computatio. SIAM, 8. [1] N. J. Higham ad L. Li. A Schur Padé algorithm for fractioal powers of a matrix. SIAM J. Matrix Aal. Appl., 32(3): , 211. [11] I. Josso ad B. Kågström. Recursive blocked algorithms for solvig triagular systems - part I: Oe-sided ad coupled Sylvester-type matrix equatios. ACM Tras. Math. Software, 28(4): , 2. [12] Numerical Algorithms Group. The NAG Fortra Library.
Blocked Schur Algorithms for Computing the Matrix Square Root. Deadman, Edvin and Higham, Nicholas J. and Ralha, Rui. MIMS EPrint: 2012.
Blocked Schur Algorithms for Computing the Matrix Square Root Deadman, Edvin and Higham, Nicholas J. and Ralha, Rui 2013 MIMS EPrint: 2012.26 Manchester Institute for Mathematical Sciences School of Mathematics
More informationBlocked Schur Algorithms for Computing the Matrix Square Root
Blocked Schur Algorithms for Computing the Matrix Square Root Edvin Deadman 1, Nicholas J. Higham 2,andRuiRalha 3 1 Numerical Algorithms Group edvin.deadman@nag.co.uk 2 University of Manchester higham@maths.manchester.ac.uk
More informationLU Decomposition Method
SOLUTION OF SIMULTANEOUS LINEAR EQUATIONS LU Decompositio Method Jamie Traha, Autar Kaw, Kevi Marti Uiversity of South Florida Uited States of America kaw@eg.usf.edu http://umericalmethods.eg.usf.edu Itroductio
More informationLecture 1: Introduction and Strassen s Algorithm
5-750: Graduate Algorithms Jauary 7, 08 Lecture : Itroductio ad Strasse s Algorithm Lecturer: Gary Miller Scribe: Robert Parker Itroductio Machie models I this class, we will primarily use the Radom Access
More informationPseudocode ( 1.1) Analysis of Algorithms. Primitive Operations. Pseudocode Details. Running Time ( 1.1) Estimating performance
Aalysis of Algorithms Iput Algorithm Output A algorithm is a step-by-step procedure for solvig a problem i a fiite amout of time. Pseudocode ( 1.1) High-level descriptio of a algorithm More structured
More informationSorting in Linear Time. Data Structures and Algorithms Andrei Bulatov
Sortig i Liear Time Data Structures ad Algorithms Adrei Bulatov Algorithms Sortig i Liear Time 7-2 Compariso Sorts The oly test that all the algorithms we have cosidered so far is compariso The oly iformatio
More informationBig-O Analysis. Asymptotics
Big-O Aalysis 1 Defiitio: Suppose that f() ad g() are oegative fuctios of. The we say that f() is O(g()) provided that there are costats C > 0 ad N > 0 such that for all > N, f() Cg(). Big-O expresses
More informationOnes Assignment Method for Solving Traveling Salesman Problem
Joural of mathematics ad computer sciece 0 (0), 58-65 Oes Assigmet Method for Solvig Travelig Salesma Problem Hadi Basirzadeh Departmet of Mathematics, Shahid Chamra Uiversity, Ahvaz, Ira Article history:
More information6.854J / J Advanced Algorithms Fall 2008
MIT OpeCourseWare http://ocw.mit.edu 6.854J / 18.415J Advaced Algorithms Fall 2008 For iformatio about citig these materials or our Terms of Use, visit: http://ocw.mit.edu/terms. 18.415/6.854 Advaced Algorithms
More informationBig-O Analysis. Asymptotics
Big-O Aalysis 1 Defiitio: Suppose that f() ad g() are oegative fuctios of. The we say that f() is O(g()) provided that there are costats C > 0 ad N > 0 such that for all > N, f() Cg(). Big-O expresses
More informationThe Closest Line to a Data Set in the Plane. David Gurney Southeastern Louisiana University Hammond, Louisiana
The Closest Lie to a Data Set i the Plae David Gurey Southeaster Louisiaa Uiversity Hammod, Louisiaa ABSTRACT This paper looks at three differet measures of distace betwee a lie ad a data set i the plae:
More informationLINPACK Benchmark Optimizations on a Virtual Processor Grid
LINPACK Bechmark Optimizatios o a Virtual Processor Grid Edward Aderso, Mayard Bradt ad Chao Yag, Cray Ic. ABSTRACT: The "massively parallel" LINPACK bechmark is a familiar measure of highperformace computig
More informationAMS subject classifications. 65F05 Direct methods for linear systems and matrix inversion
EFFICIENT SOLUTION OF A x (k) = b (k) USING A 1 ADI DITKOWSKI, GADI FIBICH, AND NIR GAVISH Abstract. I this work we cosider the problem of solvig Ax (k) = b (k), k = 1,..., K where b (k+1) = f(x (k) ).
More informationPattern Recognition Systems Lab 1 Least Mean Squares
Patter Recogitio Systems Lab 1 Least Mea Squares 1. Objectives This laboratory work itroduces the OpeCV-based framework used throughout the course. I this assigmet a lie is fitted to a set of poits usig
More informationFuzzy Minimal Solution of Dual Fully Fuzzy Matrix Equations
Iteratioal Coferece o Applied Mathematics, Simulatio ad Modellig (AMSM 2016) Fuzzy Miimal Solutio of Dual Fully Fuzzy Matrix Equatios Dequa Shag1 ad Xiaobi Guo2,* 1 Sciece Courses eachig Departmet, Gasu
More informationHigher-order iterative methods free from second derivative for solving nonlinear equations
Iteratioal Joural of the Phsical Scieces Vol 6(8, pp 887-89, 8 April, Available olie at http://wwwacademicjouralsorg/ijps DOI: 5897/IJPS45 ISSN 99-95 Academic Jourals Full Legth Research Paper Higher-order
More informationAlpha Individual Solutions MAΘ National Convention 2013
Alpha Idividual Solutios MAΘ Natioal Covetio 0 Aswers:. D. A. C 4. D 5. C 6. B 7. A 8. C 9. D 0. B. B. A. D 4. C 5. A 6. C 7. B 8. A 9. A 0. C. E. B. D 4. C 5. A 6. D 7. B 8. C 9. D 0. B TB. 570 TB. 5
More informationA New Morphological 3D Shape Decomposition: Grayscale Interframe Interpolation Method
A ew Morphological 3D Shape Decompositio: Grayscale Iterframe Iterpolatio Method D.. Vizireau Politehica Uiversity Bucharest, Romaia ae@comm.pub.ro R. M. Udrea Politehica Uiversity Bucharest, Romaia mihea@comm.pub.ro
More informationLecture 5. Counting Sort / Radix Sort
Lecture 5. Coutig Sort / Radix Sort T. H. Corme, C. E. Leiserso ad R. L. Rivest Itroductio to Algorithms, 3rd Editio, MIT Press, 2009 Sugkyukwa Uiversity Hyuseug Choo choo@skku.edu Copyright 2000-2018
More informationAn Efficient Algorithm for Graph Bisection of Triangularizations
A Efficiet Algorithm for Graph Bisectio of Triagularizatios Gerold Jäger Departmet of Computer Sciece Washigto Uiversity Campus Box 1045 Oe Brookigs Drive St. Louis, Missouri 63130-4899, USA jaegerg@cse.wustl.edu
More informationAppendix A. Use of Operators in ARPS
A Appedix A. Use of Operators i ARPS The methodology for solvig the equatios of hydrodyamics i either differetial or itegral form usig grid-poit techiques (fiite differece, fiite volume, fiite elemet)
More informationLecture 2: Spectra of Graphs
Spectral Graph Theory ad Applicatios WS 20/202 Lecture 2: Spectra of Graphs Lecturer: Thomas Sauerwald & He Su Our goal is to use the properties of the adjacecy/laplacia matrix of graphs to first uderstad
More informationLower Bounds for Sorting
Liear Sortig Topics Covered: Lower Bouds for Sortig Coutig Sort Radix Sort Bucket Sort Lower Bouds for Sortig Compariso vs. o-compariso sortig Decisio tree model Worst case lower boud Compariso Sortig
More informationAn Efficient Algorithm for Graph Bisection of Triangularizations
Applied Mathematical Scieces, Vol. 1, 2007, o. 25, 1203-1215 A Efficiet Algorithm for Graph Bisectio of Triagularizatios Gerold Jäger Departmet of Computer Sciece Washigto Uiversity Campus Box 1045, Oe
More information( n+1 2 ) , position=(7+1)/2 =4,(median is observation #4) Median=10lb
Chapter 3 Descriptive Measures Measures of Ceter (Cetral Tedecy) These measures will tell us where is the ceter of our data or where most typical value of a data set lies Mode the value that occurs most
More informationChapter 3 Classification of FFT Processor Algorithms
Chapter Classificatio of FFT Processor Algorithms The computatioal complexity of the Discrete Fourier trasform (DFT) is very high. It requires () 2 complex multiplicatios ad () complex additios [5]. As
More informationCIS 121 Data Structures and Algorithms with Java Fall Big-Oh Notation Tuesday, September 5 (Make-up Friday, September 8)
CIS 11 Data Structures ad Algorithms with Java Fall 017 Big-Oh Notatio Tuesday, September 5 (Make-up Friday, September 8) Learig Goals Review Big-Oh ad lear big/small omega/theta otatios Practice solvig
More information3D Model Retrieval Method Based on Sample Prediction
20 Iteratioal Coferece o Computer Commuicatio ad Maagemet Proc.of CSIT vol.5 (20) (20) IACSIT Press, Sigapore 3D Model Retrieval Method Based o Sample Predictio Qigche Zhag, Ya Tag* School of Computer
More informationTheory of Fuzzy Soft Matrix and its Multi Criteria in Decision Making Based on Three Basic t-norm Operators
Theory of Fuzzy Soft Matrix ad its Multi Criteria i Decisio Makig Based o Three Basic t-norm Operators Md. Jalilul Islam Modal 1, Dr. Tapa Kumar Roy 2 Research Scholar, Dept. of Mathematics, BESUS, Howrah-711103,
More informationData Structures and Algorithms. Analysis of Algorithms
Data Structures ad Algorithms Aalysis of Algorithms Outlie Ruig time Pseudo-code Big-oh otatio Big-theta otatio Big-omega otatio Asymptotic algorithm aalysis Aalysis of Algorithms Iput Algorithm Output
More informationCounting the Number of Minimum Roman Dominating Functions of a Graph
Coutig the Number of Miimum Roma Domiatig Fuctios of a Graph SHI ZHENG ad KOH KHEE MENG, Natioal Uiversity of Sigapore We provide two algorithms coutig the umber of miimum Roma domiatig fuctios of a graph
More informationThe Adjacency Matrix and The nth Eigenvalue
Spectral Graph Theory Lecture 3 The Adjacecy Matrix ad The th Eigevalue Daiel A. Spielma September 5, 2012 3.1 About these otes These otes are ot ecessarily a accurate represetatio of what happeed i class.
More informationThe golden search method: Question 1
1. Golde Sectio Search for the Mode of a Fuctio The golde search method: Questio 1 Suppose the last pair of poits at which we have a fuctio evaluatio is x(), y(). The accordig to the method, If f(x())
More information9.1. Sequences and Series. Sequences. What you should learn. Why you should learn it. Definition of Sequence
_9.qxd // : AM Page Chapter 9 Sequeces, Series, ad Probability 9. Sequeces ad Series What you should lear Use sequece otatio to write the terms of sequeces. Use factorial otatio. Use summatio otatio to
More informationLecture 18. Optimization in n dimensions
Lecture 8 Optimizatio i dimesios Itroductio We ow cosider the problem of miimizig a sigle scalar fuctio of variables, f x, where x=[ x, x,, x ]T. The D case ca be visualized as fidig the lowest poit of
More informationHow do we evaluate algorithms?
F2 Readig referece: chapter 2 + slides Algorithm complexity Big O ad big Ω To calculate ruig time Aalysis of recursive Algorithms Next time: Litterature: slides mostly The first Algorithm desig methods:
More informationNumerical Methods Lecture 6 - Curve Fitting Techniques
Numerical Methods Lecture 6 - Curve Fittig Techiques Topics motivatio iterpolatio liear regressio higher order polyomial form expoetial form Curve fittig - motivatio For root fidig, we used a give fuctio
More informationSecond-Order Domain Decomposition Method for Three-Dimensional Hyperbolic Problems
Iteratioal Mathematical Forum, Vol. 8, 013, o. 7, 311-317 Secod-Order Domai Decompositio Method for Three-Dimesioal Hyperbolic Problems Youbae Ju Departmet of Applied Mathematics Kumoh Natioal Istitute
More informationMulti-Pivot Quicksort: Theory and Experiments
Abstract Multi-Pivot Quicksort: Theory ad Experimets Shriu Kushagra skushagr@uwaterloo.ca Uiversity of Waterloo The idea of multi-pivot quicksort has recetly received the attetio of researchers after Vladimir
More informationRunning Time ( 3.1) Analysis of Algorithms. Experimental Studies. Limitations of Experiments
Ruig Time ( 3.1) Aalysis of Algorithms Iput Algorithm Output A algorithm is a step- by- step procedure for solvig a problem i a fiite amout of time. Most algorithms trasform iput objects ito output objects.
More informationAnalysis of Algorithms
Aalysis of Algorithms Iput Algorithm Output A algorithm is a step-by-step procedure for solvig a problem i a fiite amout of time. Ruig Time Most algorithms trasform iput objects ito output objects. The
More informationA Study on the Performance of Cholesky-Factorization using MPI
A Study o the Performace of Cholesky-Factorizatio usig MPI Ha S. Kim Scott B. Bade Departmet of Computer Sciece ad Egieerig Uiversity of Califoria Sa Diego {hskim, bade}@cs.ucsd.edu Abstract Cholesky-factorizatio
More informationLecture Notes 6 Introduction to algorithm analysis CSS 501 Data Structures and Object-Oriented Programming
Lecture Notes 6 Itroductio to algorithm aalysis CSS 501 Data Structures ad Object-Orieted Programmig Readig for this lecture: Carrao, Chapter 10 To be covered i this lecture: Itroductio to algorithm aalysis
More informationEE123 Digital Signal Processing
Last Time EE Digital Sigal Processig Lecture 7 Block Covolutio, Overlap ad Add, FFT Discrete Fourier Trasform Properties of the Liear covolutio through circular Today Liear covolutio with Overlap ad add
More information5.3 Recursive definitions and structural induction
/8/05 5.3 Recursive defiitios ad structural iductio CSE03 Discrete Computatioal Structures Lecture 6 A recursively defied picture Recursive defiitios e sequece of powers of is give by a = for =0,,, Ca
More informationRunning Time. Analysis of Algorithms. Experimental Studies. Limitations of Experiments
Ruig Time Aalysis of Algorithms Iput Algorithm Output A algorithm is a step-by-step procedure for solvig a problem i a fiite amout of time. Most algorithms trasform iput objects ito output objects. The
More informationMATHEMATICAL METHODS OF ANALYSIS AND EXPERIMENTAL DATA PROCESSING (Or Methods of Curve Fitting)
MATHEMATICAL METHODS OF ANALYSIS AND EXPERIMENTAL DATA PROCESSING (Or Methods of Curve Fittig) I this chapter, we will eamie some methods of aalysis ad data processig; data obtaied as a result of a give
More informationCh 9.3 Geometric Sequences and Series Lessons
Ch 9.3 Geometric Sequeces ad Series Lessos SKILLS OBJECTIVES Recogize a geometric sequece. Fid the geeral, th term of a geometric sequece. Evaluate a fiite geometric series. Evaluate a ifiite geometric
More informationSolving Fuzzy Assignment Problem Using Fourier Elimination Method
Global Joural of Pure ad Applied Mathematics. ISSN 0973-768 Volume 3, Number 2 (207), pp. 453-462 Research Idia Publicatios http://www.ripublicatio.com Solvig Fuzzy Assigmet Problem Usig Fourier Elimiatio
More informationAPPLICATION NOTE PACE1750AE BUILT-IN FUNCTIONS
APPLICATION NOTE PACE175AE BUILT-IN UNCTIONS About This Note This applicatio brief is iteded to explai ad demostrate the use of the special fuctios that are built ito the PACE175AE processor. These powerful
More informationFast Fourier Transform (FFT) Algorithms
Fast Fourier Trasform FFT Algorithms Relatio to the z-trasform elsewhere, ozero, z x z X x [ ] 2 ~ elsewhere,, ~ e j x X x x π j e z z X X π 2 ~ The DFS X represets evely spaced samples of the z- trasform
More informationEigenimages. Digital Image Processing: Bernd Girod, Stanford University -- Eigenimages 1
Eigeimages Uitary trasforms Karhue-Loève trasform ad eigeimages Sirovich ad Kirby method Eigefaces for geder recogitio Fisher liear discrimat aalysis Fisherimages ad varyig illumiatio Fisherfaces vs. eigefaces
More informationOur second algorithm. Comp 135 Machine Learning Computer Science Tufts University. Decision Trees. Decision Trees. Decision Trees.
Comp 135 Machie Learig Computer Sciece Tufts Uiversity Fall 2017 Roi Khardo Some of these slides were adapted from previous slides by Carla Brodley Our secod algorithm Let s look at a simple dataset for
More informationOrder statistics. Order Statistics. Randomized divide-andconquer. Example. CS Spring 2006
406 CS 5633 -- Sprig 006 Order Statistics Carola We Slides courtesy of Charles Leiserso with small chages by Carola We CS 5633 Aalysis of Algorithms 406 Order statistics Select the ith smallest of elemets
More informationEigenimages. Digital Image Processing: Bernd Girod, 2013 Stanford University -- Eigenimages 1
Eigeimages Uitary trasforms Karhue-Loève trasform ad eigeimages Sirovich ad Kirby method Eigefaces for geder recogitio Fisher liear discrimat aalysis Fisherimages ad varyig illumiatio Fisherfaces vs. eigefaces
More informationCSE 417: Algorithms and Computational Complexity
Time CSE 47: Algorithms ad Computatioal Readig assigmet Read Chapter of The ALGORITHM Desig Maual Aalysis & Sortig Autum 00 Paul Beame aalysis Problem size Worst-case complexity: max # steps algorithm
More informationProject 2.5 Improved Euler Implementation
Project 2.5 Improved Euler Implemetatio Figure 2.5.10 i the text lists TI-85 ad BASIC programs implemetig the improved Euler method to approximate the solutio of the iitial value problem dy dx = x+ y,
More information1.2 Binomial Coefficients and Subsets
1.2. BINOMIAL COEFFICIENTS AND SUBSETS 13 1.2 Biomial Coefficiets ad Subsets 1.2-1 The loop below is part of a program to determie the umber of triagles formed by poits i the plae. for i =1 to for j =
More informationCOMPUTER ORGANIZATION AND DESIGN The Hardware/Software Interface. Chapter 4. The Processor. Part A Datapath Design
COMPUTER ORGANIZATION AND DESIGN The Hardware/Software Iterface 5 th Editio Chapter The Processor Part A path Desig Itroductio CPU performace factors Istructio cout Determied by ISA ad compiler. CPI ad
More informationAnalysis of Algorithms
Aalysis of Algorithms Ruig Time of a algorithm Ruig Time Upper Bouds Lower Bouds Examples Mathematical facts Iput Algorithm Output A algorithm is a step-by-step procedure for solvig a problem i a fiite
More informationHomework 1 Solutions MA 522 Fall 2017
Homework 1 Solutios MA 5 Fall 017 1. Cosider the searchig problem: Iput A sequece of umbers A = [a 1,..., a ] ad a value v. Output A idex i such that v = A[i] or the special value NIL if v does ot appear
More informationA Note on Least-norm Solution of Global WireWarping
A Note o Least-orm Solutio of Global WireWarpig Charlie C. L. Wag Departmet of Mechaical ad Automatio Egieerig The Chiese Uiversity of Hog Kog Shati, N.T., Hog Kog E-mail: cwag@mae.cuhk.edu.hk Abstract
More informationProtected points in ordered trees
Applied Mathematics Letters 008 56 50 www.elsevier.com/locate/aml Protected poits i ordered trees Gi-Sag Cheo a, Louis W. Shapiro b, a Departmet of Mathematics, Sugkyukwa Uiversity, Suwo 440-746, Republic
More informationA Generalized Set Theoretic Approach for Time and Space Complexity Analysis of Algorithms and Functions
Proceedigs of the 10th WSEAS Iteratioal Coferece o APPLIED MATHEMATICS, Dallas, Texas, USA, November 1-3, 2006 316 A Geeralized Set Theoretic Approach for Time ad Space Complexity Aalysis of Algorithms
More informationOutline and Reading. Analysis of Algorithms. Running Time. Experimental Studies. Limitations of Experiments. Theoretical Analysis
Outlie ad Readig Aalysis of Algorithms Iput Algorithm Output Ruig time ( 3.) Pseudo-code ( 3.2) Coutig primitive operatios ( 3.3-3.) Asymptotic otatio ( 3.6) Asymptotic aalysis ( 3.7) Case study Aalysis
More informationElementary Educational Computer
Chapter 5 Elemetary Educatioal Computer. Geeral structure of the Elemetary Educatioal Computer (EEC) The EEC coforms to the 5 uits structure defied by vo Neuma's model (.) All uits are preseted i a simplified
More informationImproving Template Based Spike Detection
Improvig Template Based Spike Detectio Kirk Smith, Member - IEEE Portlad State Uiversity petra@ee.pdx.edu Abstract Template matchig algorithms like SSE, Covolutio ad Maximum Likelihood are well kow for
More informationA General Framework for Accurate Statistical Timing Analysis Considering Correlations
A Geeral Framework for Accurate Statistical Timig Aalysis Cosiderig Correlatios 7.4 Vishal Khadelwal Departmet of ECE Uiversity of Marylad-College Park vishalk@glue.umd.edu Akur Srivastava Departmet of
More informationTHIN LAYER ORIENTED MAGNETOSTATIC CALCULATION MODULE FOR ELMER FEM, BASED ON THE METHOD OF THE MOMENTS. Roman Szewczyk
THIN LAYER ORIENTED MAGNETOSTATIC CALCULATION MODULE FOR ELMER FEM, BASED ON THE METHOD OF THE MOMENTS Roma Szewczyk Istitute of Metrology ad Biomedical Egieerig, Warsaw Uiversity of Techology E-mail:
More informationBOOLEAN MATHEMATICS: GENERAL THEORY
CHAPTER 3 BOOLEAN MATHEMATICS: GENERAL THEORY 3.1 ISOMORPHIC PROPERTIES The ame Boolea Arithmetic was chose because it was discovered that literal Boolea Algebra could have a isomorphic umerical aspect.
More informationMath 10C Long Range Plans
Math 10C Log Rage Plas Uits: Evaluatio: Homework, projects ad assigmets 10% Uit Tests. 70% Fial Examiatio.. 20% Ay Uit Test may be rewritte for a higher mark. If the retest mark is higher, that mark will
More informationAppendix D. Controller Implementation
COMPUTER ORGANIZATION AND DESIGN The Hardware/Software Iterface 5 th Editio Appedix D Cotroller Implemetatio Cotroller Implemetatios Combiatioal logic (sigle-cycle); Fiite state machie (multi-cycle, pipelied);
More informationNew Results on Energy of Graphs of Small Order
Global Joural of Pure ad Applied Mathematics. ISSN 0973-1768 Volume 13, Number 7 (2017), pp. 2837-2848 Research Idia Publicatios http://www.ripublicatio.com New Results o Eergy of Graphs of Small Order
More informationConsider the following population data for the state of California. Year Population
Assigmets for Bradie Fall 2016 for Chapter 5 Assigmet sheet for Sectios 5.1, 5.3, 5.5, 5.6, 5.7, 5.8 Read Pages 341-349 Exercises for Sectio 5.1 Lagrage Iterpolatio #1, #4, #7, #13, #14 For #1 use MATLAB
More informationConvergence results for conditional expectations
Beroulli 11(4), 2005, 737 745 Covergece results for coditioal expectatios IRENE CRIMALDI 1 ad LUCA PRATELLI 2 1 Departmet of Mathematics, Uiversity of Bologa, Piazza di Porta Sa Doato 5, 40126 Bologa,
More informationOn Computing the Fuzzy Weighted Average Using the KM Algorithms
O Computig the Fuzzy Weighted Average Usig the KM Algorithms Feilog iu ad Jerry M Medel Sigal ad Image Processig Istitute, Departmet of Electrical Egieerig Uiversity of Souther Califoria, 3740 McClitock
More informationAnalysis Metrics. Intro to Algorithm Analysis. Slides. 12. Alg Analysis. 12. Alg Analysis
Itro to Algorithm Aalysis Aalysis Metrics Slides. Table of Cotets. Aalysis Metrics 3. Exact Aalysis Rules 4. Simple Summatio 5. Summatio Formulas 6. Order of Magitude 7. Big-O otatio 8. Big-O Theorems
More informationn n B. How many subsets of C are there of cardinality n. We are selecting elements for such a
4. [10] Usig a combiatorial argumet, prove that for 1: = 0 = Let A ad B be disjoit sets of cardiality each ad C = A B. How may subsets of C are there of cardiality. We are selectig elemets for such a subset
More informationImage Segmentation EEE 508
Image Segmetatio Objective: to determie (etract) object boudaries. It is a process of partitioig a image ito distict regios by groupig together eighborig piels based o some predefied similarity criterio.
More informationRECSY and SCASY Library Software:
Matrix computatios RECSY ad SCASY Library Software: Recursive Blocked ad Parallel Algorithms for Sylvester-type Matrix Equatios with Some Applicatios Robert raat, Isak Josso ad Bo Kågström Umeå Uiversity,
More informationAN EFFICIENT IMPLEMENTATION OF IMPLICIT OPERATOR FOR BLOCK LU-SGS METHOD
Computatioal Fluid Dyamics JOURNAL 13(): July 5 (pp.154 159) AN EFFICIENT IMPLEMENTATION OF IMPLICIT OPERATOR FOR BLOCK LU-SGS METHOD Joo Sug KIM Oh Joo KWON Abstract A efficiet implemetatio techique of
More informationc-dominating Sets for Families of Graphs
c-domiatig Sets for Families of Graphs Kelsie Syder Mathematics Uiversity of Mary Washigto April 6, 011 1 Abstract The topic of domiatio i graphs has a rich history, begiig with chess ethusiasts i the
More information. Written in factored form it is easy to see that the roots are 2, 2, i,
CMPS A Itroductio to Programmig Programmig Assigmet 4 I this assigmet you will write a java program that determies the real roots of a polyomial that lie withi a specified rage. Recall that the roots (or
More informationPerhaps the method will give that for every e > U f() > p - 3/+e There is o o-trivial upper boud for f() ad ot eve f() < Z - e. seems to be kow, where
ON MAXIMUM CHORDAL SUBGRAPH * Paul Erdos Mathematical Istitute of the Hugaria Academy of Scieces ad Reu Laskar Clemso Uiversity 1. Let G() deote a udirected graph, with vertices ad V(G) deote the vertex
More informationCOMMUNICATION-OPTIMAL PARALLEL AND SEQUENTIAL CHOLESKY DECOMPOSITION
SIA J. SCI. COUT. Vol. 2, No. 6, pp. 495 52 c 2010 Society for Idustrial ad Applied athematics COUNICATION-OTIAL ARALLEL AND SEQUENTIAL CHOLESKY DECOOSITION GREY BALLARD, JAES DEEL, OLGA HOLTZ, AND ODED
More informationSemidefinite optimization of High Performance Linpack on heterogeneous cluster
Semidefiite optimizatio of High Performace Lipack o heterogeeous cluster Draško Tomić *, Luko Gjeero** Emir Imamagić** * Hewlett-Packard Croatia, Zagreb, Croatia ** Uiversity of Zagreb Uiversity Computig
More informationA log n lower bound to compute any function in parallel Reduction and broadcast in O(log n) time Parallel prefix (scan) in O(log n) time
CS 267 Tricks with Trees Outlie A log lower boud to compute ay fuctio i parallel Reductio ad broadcast i O(log ) time Parallel prefix (sca) i O(log ) time Addig two -bit itegers i O(log ) time Multiplyig
More informationAnalysis of Algorithms
Presetatio for use with the textbook, Algorithm Desig ad Applicatios, by M. T. Goodrich ad R. Tamassia, Wiley, 2015 Aalysis of Algorithms Iput 2015 Goodrich ad Tamassia Algorithm Aalysis of Algorithms
More informationComputational Geometry
Computatioal Geometry Chapter 4 Liear programmig Duality Smallest eclosig disk O the Ageda Liear Programmig Slides courtesy of Craig Gotsma 4. 4. Liear Programmig - Example Defie: (amout amout cosumed
More informationImproved Random Graph Isomorphism
Improved Radom Graph Isomorphism Tomek Czajka Gopal Paduraga Abstract Caoical labelig of a graph cosists of assigig a uique label to each vertex such that the labels are ivariat uder isomorphism. Such
More informationAlgorithms Chapter 3 Growth of Functions
Algorithms Chapter 3 Growth of Fuctios Istructor: Chig Chi Li 林清池助理教授 chigchi.li@gmail.com Departmet of Computer Sciece ad Egieerig Natioal Taiwa Ocea Uiversity Outlie Asymptotic otatio Stadard otatios
More informationLoad balanced Parallel Prime Number Generator with Sieve of Eratosthenes on Cluster Computers *
Load balaced Parallel Prime umber Geerator with Sieve of Eratosthees o luster omputers * Soowook Hwag*, Kyusik hug**, ad Dogseug Kim* *Departmet of Electrical Egieerig Korea Uiversity Seoul, -, Rep. of
More informationAlgorithms for Disk Covering Problems with the Most Points
Algorithms for Disk Coverig Problems with the Most Poits Bi Xiao Departmet of Computig Hog Kog Polytechic Uiversity Hug Hom, Kowloo, Hog Kog csbxiao@comp.polyu.edu.hk Qigfeg Zhuge, Yi He, Zili Shao, Edwi
More informationCS 683: Advanced Design and Analysis of Algorithms
CS 683: Advaced Desig ad Aalysis of Algorithms Lecture 6, February 1, 2008 Lecturer: Joh Hopcroft Scribes: Shaomei Wu, Etha Feldma February 7, 2008 1 Threshold for k CNF Satisfiability I the previous lecture,
More informationOCR Statistics 1. Working with data. Section 3: Measures of spread
Notes ad Eamples OCR Statistics 1 Workig with data Sectio 3: Measures of spread Just as there are several differet measures of cetral tedec (averages), there are a variet of statistical measures of spread.
More informationFINITE DIFFERENCE TIME DOMAIN METHOD (FDTD)
FINIT DIFFRNC TIM DOMAIN MTOD (FDTD) The FDTD method, proposed b Yee, 1966, is aother umerical method, used widel for the solutio of M problems. It is used to solve ope-regio scatterig, radiatio, diffusio,
More information9 x and g(x) = 4. x. Find (x) 3.6. I. Combining Functions. A. From Equations. Example: Let f(x) = and its domain. Example: Let f(x) = and g(x) = x x 4
1 3.6 I. Combiig Fuctios A. From Equatios Example: Let f(x) = 9 x ad g(x) = 4 f x. Fid (x) g ad its domai. 4 Example: Let f(x) = ad g(x) = x x 4. Fid (f-g)(x) B. From Graphs: Graphical Additio. Example:
More informationIMP: Superposer Integrated Morphometrics Package Superposition Tool
IMP: Superposer Itegrated Morphometrics Package Superpositio Tool Programmig by: David Lieber ( 03) Caisius College 200 Mai St. Buffalo, NY 4208 Cocept by: H. David Sheets, Dept. of Physics, Caisius College
More informationA Variant of Pascal s Triangle Dennis Van Hise Stetson University
A Vrit of Pscl s Trigle eis V Hise tetso Uiversity I. Itroductio d Nottio It is the purpose of this pper to explore vrit of Pscl s trigle. This vrit hs the rule tht every etry, deoted s,, where,, is clculted
More informationIntro to Scientific Computing: Solutions
Itro to Scietific Computig: Solutios Dr. David M. Goulet. How may steps does it take to separate 3 objects ito groups of 4? We start with 5 objects ad apply 3 steps of the algorithm to reduce the pile
More information