Rotations in Curved Trajectories for Unconstrained Minimization
|
|
- Robert Leon Benson
- 6 years ago
- Views:
Transcription
1 Rotations in Curved rajectories for Unconstrained Minimization Alberto J Jimenez Mathematics Deartment, California Polytechnic University, San Luis Obiso, CA, USA 9407 Abstract Curved rajectories Algorithm (CA is a acage for the minimization of unconstrained functions of several variables with intervals on the variables he core algorithm is novel in that stes may follow olynomial sace curves instead of straight lines he sace curves result from truncations of a aylor series exansion of the Gradient inverse function When the series is convergent and the current guess of the solution is far, aroriate rotations of the sace curve may allow further rogress at a ste Imrovements are significant for some functions, reducing the number of Hessians required or imroving the accuracy of the solution his idea may be useful to other minimization acages Keywords Curved rajectory Minimization, Rotations of Sace Curves, Unconstrained Otimization Introduction We focus on the roblem: n Minimize f ( x, f : R R, ( n x R where f is a nonlinear continuous real scalar function, with continuous derivatives, of n variables reresented by the vector x We write the derivatives simly by: f x f x f x f x (4 (, (, (, (, etc is called the Gradient, which is is called the Hessian, which is an n n symmetric matrix he higher derivatives are tensors of higher dimension A minimum oint x * is a critical oint of the function satisfying: he first derivative f ( x an n vector function; f ( x f ( x* = 0 ( A necessary condition for a minimum is that the Hessian f ( x* must be ositive semi-definite at the critical oint (eg the Hessian has no negative eigenvalues at x *, otherwise the critical oint is not a minimum A minimum oint may not be a global minimum his aer addresses calculating a local minimum oint only yical methods for calculating a minimum start from a guess x 0 hen a direction v is selected at the th iteration ste such that the directional derivative is negative along v Now a search along this direction is made to find a suitable value of > 0 that yields f ( x + v << f ( x One can aroximate by solving the one-dimensional * Corresonding author: ajjimene@calolyedu (Alberto J Jimenez Published online at htt://journalsauborg/algorithms Coyright 0 Scientific & Academic Publishing All Rights Reserved scalar roblem, a line search: Minimize f ( x + v ( his line search can require significant comutation time and is usually unnecessary Instead, a limited number of trial values are tested until one is selected that roduces sufficient reduction in the function value his scalar minimization is a well nown roblem (see for examle [] Section 6, or [7] Chater he scalar search yields and the next guess of the solution is x + = x + v, and the iterations continue until the Gradient is as close to zero as desired, or there is lac of rogress Recent wor establishes a rust Region (see [] or [] for calculating x + using a quadratic model of the function Convergence (see for examle [8] age 5, or [7] age requires sufficient descent at each ste (in addition to other requirements, that is: f ( x+ = f ( x + v << f ( x (4 Brief Descrition of CA Algorithm A core comonent of the CA otimization acage is the addition of higher order terms to Newton's method (see [5] and [6] he higher order terms are derived by exanding the unnown minimum x* equal to the Gradient inverse function in a aylor series, and retaining u to two additional terms after the Newton term he derivation begins from ( f ( x = z x = f ( z After differentiation and much tedious algebra, the first four terms of the infinite aylor series exanded at z and evaluated at zero giving x* = ( f (0 are
2 where the vectors d, d, and d 4 are aroximated from the Hessian and Gradient at x, and from Gradients in the neighborhood of x by solving the linear systems: [ ] f ( x d = f ( x, (5 (7 he constant ε is a small ositive constant his results in an iteration, and at each ste a > 0 is selected such that x = x + + h ( satisfies (4, where h ( is a vector function reresenting the trajectory direction selected at the th ste he number of terms to use at each iteration deends on tests for convergence of the series he forms of the vector function are given by nd order trajectory h( [ ] x* = x d [ ] = d (Newton s Method rd order trajectory h( = (05 ( d d 4 th order trajectory + ( d d ( ( d ( d d 4, + 6 f ( x d = f ( x d d [ ] ( h( = 6 + d / 6 ( d d (6 4 When rd order or 4 th order is selected, the vector function h( is a olynomial curved trajectory, or sace curve, for > 0 with h(0 = 0 his algorithm converges raidly in the neighborhood of the solution when the Hessian is ositive semi-definite When far from the solution, the Hessian can be indefinite and h( may not be a descent direction for > 0 An earlier aer by this author [4] describes the factorization of the Hessian, its modification when it is not ositive definite, and storing it in sarse form In this aer we show that an aroriate rotation of the curved trajectory can often mae additional rogress when far away from the solution his rotation often trades additional Gradient and + ε [ f ( x ε d ( ε f ( x ] f x d f x d d f x d d d 6 ( 4 ( = ( ( 4 f x d d + f x ε ( ε ε ( ε ( Function evaluations to reduce the number of Hessian evaluations his tradeoff would be beneficial when the evaluation of the Hessian is very costly In some cases while there is no reduction in the number of Hessian evaluations, imrovements in the accuracy of the solution for badly scaled roblems have been observed Exloring Rotations For A wo Variable Problem We begin by focusing on a two variable roblem A rotation of the CA trajectory curve is robably best described by an angle θ, which when measured ositive counterclocwise, inserts a matrix multilying the iteration trajectory as follows: Where x = x + + R( θ h(, sinθ R( θ = sinθ Observe that when θ = 0, R(0 is the unit matrix Adding this rotation comlicates the th ste, if we were to search for the best and the best θ simultaneously Perhas this roduces the best overall results for some algorithms For this aer, since we are adding rotations to an already otimal algorithm, we first calculate the best using the algorithm as is hen we consider adding a rotation to the trajectory using the calculated to determine if greater rogress can be made towards the solution along a rotated trajectory; this can be viewed as rotating the trajectory and calculating a new value for along the rotated trajectory Before describing the calculation of a roductive angle of rotation, let us loo at an examle Consider Rosenbroc s [0] well nown banana-shaed valley function: f ( x, y 00( y x ( x = +, (8 with the usual starting guess oint, and solution: * x0 = [ x y] = [ ], x = [ ], (9 he CA acage first iteration is given by: 56 x=, 4, 0 f = f = f = ( d, d, d = = = he first iteration vector function h( is given by: ( ( h( = (06979 (
3 he CA algorithm calculates = 5, yielding: x = [ ], f = 59 Figure shows contours of f (asted from Male and this curved trajectory his figure shows Gradient vectors along the trajectory he trajectory angle with the Gradient vector lays a significant role in the imlementation of rotation as shown later Figure shows Rosenbroc s function value along this trajectory as a function of x 0 f = 4 f = 4 Original trajectory f = 4 f = 05 A rotated trajectory x x 0 f = 4 f = 4 x f = 05 Figure Contours of Rosenbroc [0] Banana Shaed Valley showing the original curved trajectory from the starting oint x 0 = (-,, and a rotated trajectory f = 4 Figure Contours of Rosenbroc [0] Banana Shaed Valley showing the first curved trajectory from the starting oint x 0 = (-, And showing x for = 5 Also shown are Gradient vector directions at some oints along the trajectory = 0, f = 4 = 5, f = 59 4 Calculating Angle Of Rotation For A wo Variable Problem After calculating the best = at the th ste with no rotation (this means θ = 0, we ursue calculating a suitable value for θ = θ that allows for further rogress along the rotated trajectory A general strategy is that calculating a roductive rotation angle should not require calculation of the Hessian he goal is to reduce the number of overall Hessians required, and using a new Hessian to determine a rotation angle would unliely hel this goal Using the Function and Gradient at = is reasonable since we need that to move on even without rotation Using an additional Function evaluation or two in the neighborhood of = also seems reasonable, but no additional Gradients will be used in the calculation of a rotation angle here are three useful angles that can be calculated We begin the derivations with the objective function as a function of θ and : f ( x = f ( θ, = f ( x + R( θ h( Figure Rosenbroc [0] Banana Shaed Valley function along the first ste trajectory for the CA algorithm as a function of he motivation for rotations is that the Gradient oints in the direction where the trajectory should be rotated to obtain smaller values of the function being minimized Given that the trajectory is a truncation of an infinite series, it is conjectured that a roductive rotation accounts for some of the excluded terms Figure shows the original trajectory, and a more roductive rotated trajectory beyond the ste oint And the artial derivative with resect to θ is given by: he above equation simlifies to: at = sin cos h ( θ θ = f R h = θ ( x x sinθ h ( = h ( h ( θ x x h ( + h ( sin θ x x And at θ = 0 and =, this artial derivative value indicates the imortance of θ :
4 = h ( h ( θ θ = 0 x x At the minimum with resect to θ, this derivative is zero, but calculating this θ * f for = 0 is not easy because the θ Gradient is changing with θ and we want to avoid more Gradient evaluations With the oor assumtion that the Gradient does not change much, we obtain a oor aroximation of θ * which we label θ r : h ( h ( x x π π θr = tan, θr h ( + h ( x x his θ r overshoots (it is larger than the actual minimum, since the Gradient terms are becoming smaller with a roductive rotation angle As shown later, we use a fraction of it as art of calculating an initial guess of a roductive θ A second useful angle is defined as θ, and this is the angle at = where the artial derivative with resect to is zero given by: sinθ h ( = f R h = ( x x sinθ h ( he above equation simlifies to: = h ( + h ( x x = h ( h ( sin θ x x At the minimum with resect to, the derivative is zero, and again with the oor assumtion that the Gradient does not change much, we obtain: h ( + h ( x x π π θ = tan, θ h ( h ( x x A third imortant angle is defined as θ G, and this is the angle between the trajectory and the Gradient at θ = 0 and = his angle is between the tangent vector to the trajectory, h (, and the Gradient he angle between two vectors is: f ih ( θg = cos, 0 θg π f h ( yically at =, the trajectory is no longer a descent direction and therefore θ G exceeds 90 degrees Recall that descending trajectories must have less than 90 degree angles with the Gradient (see for examle [9] ages 0- And the sign of θ I is equal to the sign of When the CA iteration was successful with, then we set θ = θ I Otherwise evaluate the objective function along the rotated trajectory with θ = θ I and = and then use a quadratic fit to obtain an imroved θ J When this θ J is significantly larger than the initial theta, further imrovement is obtained with another function evaluation along the rotated trajectory with θ = θ J and a second quadratic fit hus u to two function evaluations may be required to obtain a roductive theta, and some stes require none hese calculations yield the roductive rotation angle of 0885 radians shown in Figure for Rosenbroc function with no additional function samles he angles described above (in radians are: θ r = 0707, θ = and θ π / = he artial derivative is G Figures 4 and 5 show the rotation angle contours and a D grah for Rosenbroc s Banana function versus θ and Figure 4 Contours of the rotation angle θ for Rosenbroc [0] Banana Shaed Valley function along the first ste trajectory for the CA algorithm as a function of f θ θ = 0 = 859 θ = 0 θ = 0 θ = 0 θ = 0 θ θ = 0 With these three angles, we select an initial guess for θ to be the following: π π θi = 04max( θg, θ, min( θr, θr
5 f x 0 f = 4 f = 4 Current oint with = 5 on original trajectory he new better oint on rotated trajectory is with = 67 x f = 05 f = 4 Closest oint on rotated trajectory is with = 544 Figure 5 A D grah of Rosenbroc [0] Banana Shaed Valley function along the first ste trajectory for the CA algorithm as functions of and θ Figure 6 Contours of Rosenbroc [0] Banana Shaed Valley showing the original curved trajectory and rotated trajectory and starting guess to search for a better oint leading to = 67 After imlementing rotations in the CA acage, and testing hundreds of functions, it was determined that rotations should not be attemted unless the following is satisfied: CA trajectory was the result of a converging series (successive terms have smaller norm, Hessian is ositive definite when the order selected is two (Newton's direction, Current oint is far from the solution (Gradient norm is greater than, CA s th ste wored well ( was not too small, is not too small (greater than 0005 times Gradient norm, θ = θ= 0 Calculated θ is not too small (greater than 000, Function value on rotated trajectory is smaller than on the original trajectory he last item to discuss is the search for a new = along the rotated trajectory We want to start searching at the value on the trajectory which corresonds to the oint closest to the oint for on the non-rotated trajectory his is done by calculating smallest distance between the two trajectories without needing any objective function evaluations Figure 6 shows the results for Rosenbroc s function 5 More han wo Variable Problems Including more variables into rotating the trajectory starts by determining the most imortant coordinates to include in a rotation calculation We argue that they are the coordinates that imact the Gradient angle with the trajectory the most his means sorting the dot roduct terms of f h ( i Initially we only focused on the terms that are ositive in the dot roduct thining those coordinates are the ones maing the angle the largest However after exerimentation, sorting on the absolute value of the dot roduct terms roved to be the best determinant of the most imortant coordinates Also exerimentally we determined that terms that are smaller than 5% of the largest one are not worth considering, and no more than the to 4 imortant coordinates needed to be considered After raning the coordinates, the to two coordinates are used to calculate the first angle of rotation θ 0, as exlained in the revious section he only change to the revious section is the calculation of θ I when there are three imortant coordinates: θ = max(05max( θ π /, θ, 09min( θ, π / θ, I G r r and when there are four imortant coordinates: θ = max(06max( θ π /, θ, 08min( θ, π / θ, I G r r where the factors were otimized exerimentally hen all the imortant coordinates are added to the rotation as follows Each of these coordinates in order is aired one at a time, with all the receding ones, and the angle of rotation is calculated as described above For examle, the rotation matrix for a three variable roblem where all are imortant variables is effectively the roduct of these three rotation matrices: sinθ 0 sinθ0 0 R( θ = 0 sinθ 0 0 sinθ sinθ sinθ 0 0 0
6 he imlementation always handles rotation between two variables at a time; and in fact, it never oerates with any matrix larger than 6 Results Of Some Cuter Unconstrained Problems [] he Curved rajectories Algorithm (CA acage, including the otion of rotation, has been imlemented to run in the Cuter environment, and is available from the author for academic use he CA acage solves all the error-free unconstrained roblems, including ones with interval constraints on the variables able shows details on some roblems running with and without rotation Stoing criteria has been selected to achieve solutions as good as ublished results, tyically with Gradient norms less than able Summary of CA acage erformance on a samle of small unconstrained roblems from Cuter Column labels: Rot is number of iterations that used rotations; It is number of iterations, NF, NG, and NH are the number of Function, Gradient and Hessian evaluations; f and Gradient columns are the ending Function and Gradient norm values Problem size ROSENBR WOOD 4 BRYBND 0 CHNROSNB 0 DIXMAANJ 5 ENGVAL GROWHLS HEAR6LS 6 HEAR8LS 8 MARAOSB MEYER PFILS PFILS PFILS PFI4LS YFIU Rot It NF NG NH f Gradient norm E-4 7E E- 779E E-0 48E E-9 4E E-06 0E E-06 74E E-08 06E E- 0E E E E- 54E E- 40E E E E-04 4E E-0 48E E-4 65E E-6 669E E E E E E-04 4E E-04 8E E-06 08E E-05 49E E-06 0E E E E-07 8E E-06 47E E- 75E E- 4E-04 able summarizes the results on 7 roblems with less than 00 variables, giving a view of the liely imact of rotations for a large number of difficult roblems For sure, many of the roblems do not use rotations Running times are about the same Rotations tyically save Hessian evaluations at the exense of additional Gradient and Function evaluations his would be attractive when evaluating the Hessian is very costly able otals for 7 unconstrained Cuter roblems with less than 00 variables Pacage Iterations NF NG NH CA CA with rotation As the samles in the ables show, rotation maes significant imrovements for some roblems, no change to other roblems, and causes a few roblems to tae a longer ath to the solution; this is not an unexected result with many difficult roblems In many cases rotations imrove the solution found he singular and oorly scaled roblem SCOSINE in able is worth highlighting: rotations ended u needing one more Hessian, but the solution is more accurate able Summary of CA acage erformance on a samle of large unconstrained roblems from Cuter Column labels: Rot is number of iterations that used rotations; It is number of iterations, NF, NG, and NH are the number of Function, Gradient and Hessian evaluations; f and Gradient columns are the ending Function and Gradient norm values Problem size CHAINWOO,000 SBRYBND,000 SCOSINE 0,000 7 Conclusion Rot It NF NG NH f Gradient norm E E E-9 58E E- 05E E+0 46E E+0 799E-07 his aer resents the derivation and imlementation of adding a rotation otion to a minimization algorithm It has been imlemented in the already very efficient CA acage and roduced encouraging results, though in a few cases it caused more wor Using rotations changes the ath taen to arrive at the solution and this can sometimes end u in a region that is not smooth causing more wor to obtain the solution Rotations reduced the wor for the vast majority of the roblems and saved Hessian evaluations, and this would be of benefit when the Hessian is costly to evaluate
7 REFERENCES [] A R Conn, N I M Gould and P L oint, rust-region Methods, SIAM, 600 University City Science Center, Philadelhia, PA, 000 [] N I M Gould, D Orban and P L oint, CUEr and SifDec: A constrained and unconstrained testing environment, revisited, ACM ransactions on Mathematical Software, 9, 4 (00, 7 94 [] D M Himmelblau, Alied Nonlinear Programming, McGraw-Hill Boo Comany, NY, 97 [4] A J Jiménez, Sarse Hessian factorization in curved trajectories for unconstrained minimization, Otimization Methods and Software, London DOI:0080/ , 0 [5] A J Jiménez, A Variable-Order Nonlinear Programming Algorithm for Use in Comuter-Aided Circuit Design and Analysis, PhD Dissertation, University of Florida, Gainesville, Florida, 976 Available at the University of Florida, George A Smathers Libraries, htt://archiveorg/details/variableordernon00jim [6] AJ Jiménez and SW Director, A Variable-Order Algorithm for Unconstrained Minimization, in Proceedings 976 IEEE International Symosium on Circuit and Systems, Aril 976 [7] D Kincaid and W Cheney, Numerical Analysis Mathematics of Scientific Comuting, hird Edition, Brooc/Cole, Pacific Grove, Ca, 00 [8] D G Luenberger, Introduction to Linear and Nonlinear Programming, Addison-Wesley Publishing Comany, Menlo Par, CA, 965 [9] J Nocedal and S J Wright, Numerical Otimization, nd edition, Sriger, NY, NY, 006 [0] H H Rosenbroc, An automatic method for finding the greatest or the least value of a function, Comuter Journal, (960, [] D C Sorensen, Newton s method with a model trust region modification, SIAM Journal on Numerical Analysis, 9 (98,
Sparse Hessian Factorization in Curved Trajectories for Minimization
Sparse Hessian Factorization in Curved Trajectories for Minimization ALBERTO J JIMÉNEZ * The Curved Trajectories Algorithm (CTA) for the minimization of unconstrained functions of several variables follows
More informationFeedback-error control
Chater 4 Feedback-error control 4.1 Introduction This chater exlains the feedback-error (FBE) control scheme originally described by Kawato [, 87, 8]. FBE is a widely used neural network based controller
More information4. Score normalization technical details We now discuss the technical details of the score normalization method.
SMT SCORING SYSTEM This document describes the scoring system for the Stanford Math Tournament We begin by giving an overview of the changes to scoring and a non-technical descrition of the scoring rules
More informationNumerical Methods for Particle Tracing in Vector Fields
On-Line Visualization Notes Numerical Methods for Particle Tracing in Vector Fields Kenneth I. Joy Visualization and Grahics Research Laboratory Deartment of Comuter Science University of California, Davis
More informationJohn Weatherwax. Analysis of Parallel Depth First Search Algorithms
Sulementary Discussions and Solutions to Selected Problems in: Introduction to Parallel Comuting by Viin Kumar, Ananth Grama, Anshul Guta, & George Karyis John Weatherwax Chater 8 Analysis of Parallel
More informationFor q 0; 1; : : : ; `? 1, we have m 0; 1; : : : ; q? 1. The set fh j(x) : j 0; 1; ; : : : ; `? 1g forms a basis for the tness functions dened on the i
Comuting with Haar Functions Sami Khuri Deartment of Mathematics and Comuter Science San Jose State University One Washington Square San Jose, CA 9519-0103, USA khuri@juiter.sjsu.edu Fax: (40)94-500 Keywords:
More informationPrincipal Components Analysis and Unsupervised Hebbian Learning
Princial Comonents Analysis and Unsuervised Hebbian Learning Robert Jacobs Deartment of Brain & Cognitive Sciences University of Rochester Rochester, NY 1467, USA August 8, 008 Reference: Much of the material
More informationNONLINEAR OPTIMIZATION WITH CONVEX CONSTRAINTS. The Goldstein-Levitin-Polyak algorithm
- (23) NLP - NONLINEAR OPTIMIZATION WITH CONVEX CONSTRAINTS The Goldstein-Levitin-Polya algorithm We consider an algorithm for solving the otimization roblem under convex constraints. Although the convexity
More informationApproximating min-max k-clustering
Aroximating min-max k-clustering Asaf Levin July 24, 2007 Abstract We consider the roblems of set artitioning into k clusters with minimum total cost and minimum of the maximum cost of a cluster. The cost
More information216 S. Chandrasearan and I.C.F. Isen Our results dier from those of Sun [14] in two asects: we assume that comuted eigenvalues or singular values are
Numer. Math. 68: 215{223 (1994) Numerische Mathemati c Sringer-Verlag 1994 Electronic Edition Bacward errors for eigenvalue and singular value decomositions? S. Chandrasearan??, I.C.F. Isen??? Deartment
More informationPreconditioning techniques for Newton s method for the incompressible Navier Stokes equations
Preconditioning techniques for Newton s method for the incomressible Navier Stokes equations H. C. ELMAN 1, D. LOGHIN 2 and A. J. WATHEN 3 1 Deartment of Comuter Science, University of Maryland, College
More informationRecursive Estimation of the Preisach Density function for a Smart Actuator
Recursive Estimation of the Preisach Density function for a Smart Actuator Ram V. Iyer Deartment of Mathematics and Statistics, Texas Tech University, Lubbock, TX 7949-142. ABSTRACT The Preisach oerator
More informationLilian Markenzon 1, Nair Maria Maia de Abreu 2* and Luciana Lee 3
Pesquisa Oeracional (2013) 33(1): 123-132 2013 Brazilian Oerations Research Society Printed version ISSN 0101-7438 / Online version ISSN 1678-5142 www.scielo.br/oe SOME RESULTS ABOUT THE CONNECTIVITY OF
More informationA Network-Flow Based Cell Sizing Algorithm
A Networ-Flow Based Cell Sizing Algorithm Huan Ren and Shantanu Dutt Det. of ECE, University of Illinois-Chicago Abstract We roose a networ flow based algorithm for the area-constrained timing-driven discrete
More informationREFINED STRAIN ENERGY OF THE SHELL
REFINED STRAIN ENERGY OF THE SHELL Ryszard A. Walentyński Deartment of Building Structures Theory, Silesian University of Technology, Gliwice, PL44-11, Poland ABSTRACT The aer rovides information on evaluation
More informationAn Investigation on the Numerical Ill-conditioning of Hybrid State Estimators
An Investigation on the Numerical Ill-conditioning of Hybrid State Estimators S. K. Mallik, Student Member, IEEE, S. Chakrabarti, Senior Member, IEEE, S. N. Singh, Senior Member, IEEE Deartment of Electrical
More informationParticipation Factors. However, it does not give the influence of each state on the mode.
Particiation Factors he mode shae, as indicated by the right eigenvector, gives the relative hase of each state in a articular mode. However, it does not give the influence of each state on the mode. We
More informationSums of independent random variables
3 Sums of indeendent random variables This lecture collects a number of estimates for sums of indeendent random variables with values in a Banach sace E. We concentrate on sums of the form N γ nx n, where
More informationESTIMATION OF THE OUTPUT DEVIATION NORM FOR UNCERTAIN, DISCRETE-TIME NONLINEAR SYSTEMS IN A STATE DEPENDENT FORM
Int. J. Al. Math. Comut. Sci. 2007 Vol. 17 No. 4 505 513 DOI: 10.2478/v10006-007-0042-z ESTIMATION OF THE OUTPUT DEVIATION NORM FOR UNCERTAIN DISCRETE-TIME NONLINEAR SYSTEMS IN A STATE DEPENDENT FORM PRZEMYSŁAW
More informationControllability and Resiliency Analysis in Heat Exchanger Networks
609 A ublication of CHEMICAL ENGINEERING RANSACIONS VOL. 6, 07 Guest Editors: Petar S Varbanov, Rongxin Su, Hon Loong Lam, Xia Liu, Jiří J Klemeš Coyright 07, AIDIC Servizi S.r.l. ISBN 978-88-95608-5-8;
More informationPretest (Optional) Use as an additional pacing tool to guide instruction. August 21
Trimester 1 Pretest (Otional) Use as an additional acing tool to guide instruction. August 21 Beyond the Basic Facts In Trimester 1, Grade 8 focus on multilication. Daily Unit 1: Rational vs. Irrational
More informationc Copyright by Helen J. Elwood December, 2011
c Coyright by Helen J. Elwood December, 2011 CONSTRUCTING COMPLEX EQUIANGULAR PARSEVAL FRAMES A Dissertation Presented to the Faculty of the Deartment of Mathematics University of Houston In Partial Fulfillment
More informationUnit 1 - Computer Arithmetic
FIXD-POINT (FX) ARITHMTIC Unit 1 - Comuter Arithmetic INTGR NUMBRS n bit number: b n 1 b n 2 b 0 Decimal Value Range of values UNSIGND n 1 SIGND D = b i 2 i D = 2 n 1 b n 1 + b i 2 i n 2 i=0 i=0 [0, 2
More informationFinding Shortest Hamiltonian Path is in P. Abstract
Finding Shortest Hamiltonian Path is in P Dhananay P. Mehendale Sir Parashurambhau College, Tilak Road, Pune, India bstract The roblem of finding shortest Hamiltonian ath in a eighted comlete grah belongs
More informationThe University of the State of New York REGENTS HIGH SCHOOL EXAMINATION COURSE III. Wednesday, August 16, :30 to 11:30 a.m.
The University of the State of New York REGENTS HIGH SCHOOL EXAMINATION THREE-YEAR SEQUENCE FOR HIGH SCHOOL MATHEMATICS COURSE III Wednesday, August 6, 000 8:0 to :0 a.m., only Notice... Scientific calculators
More informationElements of Asymptotic Theory. James L. Powell Department of Economics University of California, Berkeley
Elements of Asymtotic Theory James L. Powell Deartment of Economics University of California, Berkeley Objectives of Asymtotic Theory While exact results are available for, say, the distribution of the
More informationNumerical Linear Algebra
Numerical Linear Algebra Numerous alications in statistics, articularly in the fitting of linear models. Notation and conventions: Elements of a matrix A are denoted by a ij, where i indexes the rows and
More informationGOOD MODELS FOR CUBIC SURFACES. 1. Introduction
GOOD MODELS FOR CUBIC SURFACES ANDREAS-STEPHAN ELSENHANS Abstract. This article describes an algorithm for finding a model of a hyersurface with small coefficients. It is shown that the aroach works in
More informationRANDOM WALKS AND PERCOLATION: AN ANALYSIS OF CURRENT RESEARCH ON MODELING NATURAL PROCESSES
RANDOM WALKS AND PERCOLATION: AN ANALYSIS OF CURRENT RESEARCH ON MODELING NATURAL PROCESSES AARON ZWIEBACH Abstract. In this aer we will analyze research that has been recently done in the field of discrete
More informationRound-off Errors and Computer Arithmetic - (1.2)
Round-off Errors and Comuter Arithmetic - (.). Round-off Errors: Round-off errors is roduced when a calculator or comuter is used to erform real number calculations. That is because the arithmetic erformed
More informationChapter 7 Rational and Irrational Numbers
Chater 7 Rational and Irrational Numbers In this chater we first review the real line model for numbers, as discussed in Chater 2 of seventh grade, by recalling how the integers and then the rational numbers
More informationUse of Transformations and the Repeated Statement in PROC GLM in SAS Ed Stanek
Use of Transformations and the Reeated Statement in PROC GLM in SAS Ed Stanek Introduction We describe how the Reeated Statement in PROC GLM in SAS transforms the data to rovide tests of hyotheses of interest.
More informationRobust Predictive Control of Input Constraints and Interference Suppression for Semi-Trailer System
Vol.7, No.7 (4),.37-38 htt://dx.doi.org/.457/ica.4.7.7.3 Robust Predictive Control of Inut Constraints and Interference Suression for Semi-Trailer System Zhao, Yang Electronic and Information Technology
More informationImplementation and Validation of Finite Volume C++ Codes for Plane Stress Analysis
CST0 191 October, 011, Krabi Imlementation and Validation of Finite Volume C++ Codes for Plane Stress Analysis Chakrit Suvanjumrat and Ekachai Chaichanasiri* Deartment of Mechanical Engineering, Faculty
More informationUniformly best wavenumber approximations by spatial central difference operators: An initial investigation
Uniformly best wavenumber aroximations by satial central difference oerators: An initial investigation Vitor Linders and Jan Nordström Abstract A characterisation theorem for best uniform wavenumber aroximations
More informationCombining Logistic Regression with Kriging for Mapping the Risk of Occurrence of Unexploded Ordnance (UXO)
Combining Logistic Regression with Kriging for Maing the Risk of Occurrence of Unexloded Ordnance (UXO) H. Saito (), P. Goovaerts (), S. A. McKenna (2) Environmental and Water Resources Engineering, Deartment
More informationSystem Reliability Estimation and Confidence Regions from Subsystem and Full System Tests
009 American Control Conference Hyatt Regency Riverfront, St. Louis, MO, USA June 0-, 009 FrB4. System Reliability Estimation and Confidence Regions from Subsystem and Full System Tests James C. Sall Abstract
More informationMetrics Performance Evaluation: Application to Face Recognition
Metrics Performance Evaluation: Alication to Face Recognition Naser Zaeri, Abeer AlSadeq, and Abdallah Cherri Electrical Engineering Det., Kuwait University, P.O. Box 5969, Safat 6, Kuwait {zaery, abeer,
More informationAlmost 4000 years ago, Babylonians had discovered the following approximation to. x 2 dy 2 =1, (5.0.2)
Chater 5 Pell s Equation One of the earliest issues graled with in number theory is the fact that geometric quantities are often not rational. For instance, if we take a right triangle with two side lengths
More information1 Extremum Estimators
FINC 9311-21 Financial Econometrics Handout Jialin Yu 1 Extremum Estimators Let θ 0 be a vector of k 1 unknown arameters. Extremum estimators: estimators obtained by maximizing or minimizing some objective
More informationSpectral Clustering based on the graph p-laplacian
Sectral Clustering based on the grah -Lalacian Thomas Bühler tb@cs.uni-sb.de Matthias Hein hein@cs.uni-sb.de Saarland University Comuter Science Deartment Camus E 663 Saarbrücken Germany Abstract We resent
More informationSome results of convex programming complexity
2012c12 $ Ê Æ Æ 116ò 14Ï Dec., 2012 Oerations Research Transactions Vol.16 No.4 Some results of convex rogramming comlexity LOU Ye 1,2 GAO Yuetian 1 Abstract Recently a number of aers were written that
More informationOn Fractional Predictive PID Controller Design Method Emmanuel Edet*. Reza Katebi.**
On Fractional Predictive PID Controller Design Method Emmanuel Edet*. Reza Katebi.** * echnology and Innovation Centre, Level 4, Deartment of Electronic and Electrical Engineering, University of Strathclyde,
More informationObserver/Kalman Filter Time Varying System Identification
Observer/Kalman Filter Time Varying System Identification Manoranjan Majji Texas A&M University, College Station, Texas, USA Jer-Nan Juang 2 National Cheng Kung University, Tainan, Taiwan and John L. Junins
More informationA Parallel Algorithm for Minimization of Finite Automata
A Parallel Algorithm for Minimization of Finite Automata B. Ravikumar X. Xiong Deartment of Comuter Science University of Rhode Island Kingston, RI 02881 E-mail: fravi,xiongg@cs.uri.edu Abstract In this
More informationSeries Handout A. 1. Determine which of the following sums are geometric. If the sum is geometric, express the sum in closed form.
Series Handout A. Determine which of the following sums are geometric. If the sum is geometric, exress the sum in closed form. 70 a) k= ( k ) b) 50 k= ( k )2 c) 60 k= ( k )k d) 60 k= (.0)k/3 2. Find the
More informationSTABILITY ANALYSIS TOOL FOR TUNING UNCONSTRAINED DECENTRALIZED MODEL PREDICTIVE CONTROLLERS
STABILITY ANALYSIS TOOL FOR TUNING UNCONSTRAINED DECENTRALIZED MODEL PREDICTIVE CONTROLLERS Massimo Vaccarini Sauro Longhi M. Reza Katebi D.I.I.G.A., Università Politecnica delle Marche, Ancona, Italy
More informationEstimating Time-Series Models
Estimating ime-series Models he Box-Jenkins methodology for tting a model to a scalar time series fx t g consists of ve stes:. Decide on the order of di erencing d that is needed to roduce a stationary
More informationResearch Article An iterative Algorithm for Hemicontractive Mappings in Banach Spaces
Abstract and Alied Analysis Volume 2012, Article ID 264103, 11 ages doi:10.1155/2012/264103 Research Article An iterative Algorithm for Hemicontractive Maings in Banach Saces Youli Yu, 1 Zhitao Wu, 2 and
More informationSection 0.10: Complex Numbers from Precalculus Prerequisites a.k.a. Chapter 0 by Carl Stitz, PhD, and Jeff Zeager, PhD, is available under a Creative
Section 0.0: Comlex Numbers from Precalculus Prerequisites a.k.a. Chater 0 by Carl Stitz, PhD, and Jeff Zeager, PhD, is available under a Creative Commons Attribution-NonCommercial-ShareAlike.0 license.
More informationA Closed-Form Solution to the Minimum V 2
Celestial Mechanics and Dynamical Astronomy manuscrit No. (will be inserted by the editor) Martín Avendaño Daniele Mortari A Closed-Form Solution to the Minimum V tot Lambert s Problem Received: Month
More informationDesign of NARMA L-2 Control of Nonlinear Inverted Pendulum
International Research Journal of Alied and Basic Sciences 016 Available online at www.irjabs.com ISSN 51-838X / Vol, 10 (6): 679-684 Science Exlorer Publications Design of NARMA L- Control of Nonlinear
More information2-D Analysis for Iterative Learning Controller for Discrete-Time Systems With Variable Initial Conditions Yong FANG 1, and Tommy W. S.
-D Analysis for Iterative Learning Controller for Discrete-ime Systems With Variable Initial Conditions Yong FANG, and ommy W. S. Chow Abstract In this aer, an iterative learning controller alying to linear
More informationMODEL-BASED MULTIPLE FAULT DETECTION AND ISOLATION FOR NONLINEAR SYSTEMS
MODEL-BASED MULIPLE FAUL DEECION AND ISOLAION FOR NONLINEAR SYSEMS Ivan Castillo, and homas F. Edgar he University of exas at Austin Austin, X 78712 David Hill Chemstations Houston, X 77009 Abstract A
More informationParallelism and Locality in Priority Queues. A. Ranade S. Cheng E. Deprit J. Jones S. Shih. University of California. Berkeley, CA 94720
Parallelism and Locality in Priority Queues A. Ranade S. Cheng E. Derit J. Jones S. Shih Comuter Science Division University of California Berkeley, CA 94720 Abstract We exlore two ways of incororating
More informationA numerical implementation of a predictor-corrector algorithm for sufcient linear complementarity problem
A numerical imlementation of a redictor-corrector algorithm for sufcient linear comlementarity roblem BENTERKI DJAMEL University Ferhat Abbas of Setif-1 Faculty of science Laboratory of fundamental and
More informationMonopolist s mark-up and the elasticity of substitution
Croatian Oerational Research Review 377 CRORR 8(7), 377 39 Monoolist s mark-u and the elasticity of substitution Ilko Vrankić, Mira Kran, and Tomislav Herceg Deartment of Economic Theory, Faculty of Economics
More informationCOMPARISON OF VARIOUS OPTIMIZATION TECHNIQUES FOR DESIGN FIR DIGITAL FILTERS
NCCI 1 -National Conference on Comutational Instrumentation CSIO Chandigarh, INDIA, 19- March 1 COMPARISON OF VARIOUS OPIMIZAION ECHNIQUES FOR DESIGN FIR DIGIAL FILERS Amanjeet Panghal 1, Nitin Mittal,Devender
More informationA Social Welfare Optimal Sequential Allocation Procedure
A Social Welfare Otimal Sequential Allocation Procedure Thomas Kalinowsi Universität Rostoc, Germany Nina Narodytsa and Toby Walsh NICTA and UNSW, Australia May 2, 201 Abstract We consider a simle sequential
More informationMODELING THE RELIABILITY OF C4ISR SYSTEMS HARDWARE/SOFTWARE COMPONENTS USING AN IMPROVED MARKOV MODEL
Technical Sciences and Alied Mathematics MODELING THE RELIABILITY OF CISR SYSTEMS HARDWARE/SOFTWARE COMPONENTS USING AN IMPROVED MARKOV MODEL Cezar VASILESCU Regional Deartment of Defense Resources Management
More informationChurilova Maria Saint-Petersburg State Polytechnical University Department of Applied Mathematics
Churilova Maria Saint-Petersburg State Polytechnical University Deartment of Alied Mathematics Technology of EHIS (staming) alied to roduction of automotive arts The roblem described in this reort originated
More informationEfficient & Robust LK for Mobile Vision
Efficient & Robust LK for Mobile Vision Instructor - Simon Lucey 16-623 - Designing Comuter Vision As Direct Method (ours) Indirect Method (ORB+RANSAC) H. Alismail, B. Browning, S. Lucey Bit-Planes: Dense
More informationFE FORMULATIONS FOR PLASTICITY
G These slides are designed based on the book: Finite Elements in Plasticity Theory and Practice, D.R.J. Owen and E. Hinton, 1970, Pineridge Press Ltd., Swansea, UK. 1 Course Content: A INTRODUCTION AND
More informationComputer arithmetic. Intensive Computation. Annalisa Massini 2017/2018
Comuter arithmetic Intensive Comutation Annalisa Massini 7/8 Intensive Comutation - 7/8 References Comuter Architecture - A Quantitative Aroach Hennessy Patterson Aendix J Intensive Comutation - 7/8 3
More informationOn Line Parameter Estimation of Electric Systems using the Bacterial Foraging Algorithm
On Line Parameter Estimation of Electric Systems using the Bacterial Foraging Algorithm Gabriel Noriega, José Restreo, Víctor Guzmán, Maribel Giménez and José Aller Universidad Simón Bolívar Valle de Sartenejas,
More informationVariable Selection and Model Building
LINEAR REGRESSION ANALYSIS MODULE XIII Lecture - 38 Variable Selection and Model Building Dr. Shalabh Deartment of Mathematics and Statistics Indian Institute of Technology Kanur Evaluation of subset regression
More informationA Simple Weight Decay Can Improve. Abstract. It has been observed in numerical simulations that a weight decay can improve
In Advances in Neural Information Processing Systems 4, J.E. Moody, S.J. Hanson and R.P. Limann, eds. Morgan Kaumann Publishers, San Mateo CA, 1995,. 950{957. A Simle Weight Decay Can Imrove Generalization
More information2 Asymptotic density and Dirichlet density
8.785: Analytic Number Theory, MIT, sring 2007 (K.S. Kedlaya) Primes in arithmetic rogressions In this unit, we first rove Dirichlet s theorem on rimes in arithmetic rogressions. We then rove the rime
More informationMATHEMATICAL MODELLING OF THE WIRELESS COMMUNICATION NETWORK
Comuter Modelling and ew Technologies, 5, Vol.9, o., 3-39 Transort and Telecommunication Institute, Lomonosov, LV-9, Riga, Latvia MATHEMATICAL MODELLIG OF THE WIRELESS COMMUICATIO ETWORK M. KOPEETSK Deartment
More informationNUMERICAL AND THEORETICAL INVESTIGATIONS ON DETONATION- INERT CONFINEMENT INTERACTIONS
NUMERICAL AND THEORETICAL INVESTIGATIONS ON DETONATION- INERT CONFINEMENT INTERACTIONS Tariq D. Aslam and John B. Bdzil Los Alamos National Laboratory Los Alamos, NM 87545 hone: 1-55-667-1367, fax: 1-55-667-6372
More informationAnalysis of some entrance probabilities for killed birth-death processes
Analysis of some entrance robabilities for killed birth-death rocesses Master s Thesis O.J.G. van der Velde Suervisor: Dr. F.M. Sieksma July 5, 207 Mathematical Institute, Leiden University Contents Introduction
More informationRecent Developments in Multilayer Perceptron Neural Networks
Recent Develoments in Multilayer Percetron eural etworks Walter H. Delashmit Lockheed Martin Missiles and Fire Control Dallas, Texas 75265 walter.delashmit@lmco.com walter.delashmit@verizon.net Michael
More informationModeling and Estimation of Full-Chip Leakage Current Considering Within-Die Correlation
6.3 Modeling and Estimation of Full-Chi Leaage Current Considering Within-Die Correlation Khaled R. eloue, Navid Azizi, Farid N. Najm Deartment of ECE, University of Toronto,Toronto, Ontario, Canada {haled,nazizi,najm}@eecg.utoronto.ca
More information2 Asymptotic density and Dirichlet density
8.785: Analytic Number Theory, MIT, sring 2007 (K.S. Kedlaya) Primes in arithmetic rogressions In this unit, we first rove Dirichlet s theorem on rimes in arithmetic rogressions. We then rove the rime
More informationHidden Predictors: A Factor Analysis Primer
Hidden Predictors: A Factor Analysis Primer Ryan C Sanchez Western Washington University Factor Analysis is a owerful statistical method in the modern research sychologist s toolbag When used roerly, factor
More informationSolvability and Number of Roots of Bi-Quadratic Equations over p adic Fields
Malaysian Journal of Mathematical Sciences 10(S February: 15-35 (016 Secial Issue: The 3 rd International Conference on Mathematical Alications in Engineering 014 (ICMAE 14 MALAYSIAN JOURNAL OF MATHEMATICAL
More informationA Recursive Block Incomplete Factorization. Preconditioner for Adaptive Filtering Problem
Alied Mathematical Sciences, Vol. 7, 03, no. 63, 3-3 HIKARI Ltd, www.m-hiari.com A Recursive Bloc Incomlete Factorization Preconditioner for Adative Filtering Problem Shazia Javed School of Mathematical
More informationCR extensions with a classical Several Complex Variables point of view. August Peter Brådalen Sonne Master s Thesis, Spring 2018
CR extensions with a classical Several Comlex Variables oint of view August Peter Brådalen Sonne Master s Thesis, Sring 2018 This master s thesis is submitted under the master s rogramme Mathematics, with
More informationUncorrelated Multilinear Principal Component Analysis for Unsupervised Multilinear Subspace Learning
TNN-2009-P-1186.R2 1 Uncorrelated Multilinear Princial Comonent Analysis for Unsuervised Multilinear Subsace Learning Haiing Lu, K. N. Plataniotis and A. N. Venetsanooulos The Edward S. Rogers Sr. Deartment
More informationGeneration of Linear Models using Simulation Results
4. IMACS-Symosium MATHMOD, Wien, 5..003,. 436-443 Generation of Linear Models using Simulation Results Georg Otte, Sven Reitz, Joachim Haase Fraunhofer Institute for Integrated Circuits, Branch Lab Design
More informationUsing the Divergence Information Criterion for the Determination of the Order of an Autoregressive Process
Using the Divergence Information Criterion for the Determination of the Order of an Autoregressive Process P. Mantalos a1, K. Mattheou b, A. Karagrigoriou b a.deartment of Statistics University of Lund
More informationI have not proofread these notes; so please watch out for typos, anything misleading or just plain wrong.
hermodynamics I have not roofread these notes; so lease watch out for tyos, anything misleading or just lain wrong. Please read ages 227 246 in Chater 8 of Kittel and Kroemer and ay attention to the first
More informationAn Analysis of Reliable Classifiers through ROC Isometrics
An Analysis of Reliable Classifiers through ROC Isometrics Stijn Vanderlooy s.vanderlooy@cs.unimaas.nl Ida G. Srinkhuizen-Kuyer kuyer@cs.unimaas.nl Evgueni N. Smirnov smirnov@cs.unimaas.nl MICC-IKAT, Universiteit
More informationEE 508 Lecture 13. Statistical Characterization of Filter Characteristics
EE 508 Lecture 3 Statistical Characterization of Filter Characteristics Comonents used to build filters are not recisely redictable L C Temerature Variations Manufacturing Variations Aging Model variations
More informationNetwork Configuration Control Via Connectivity Graph Processes
Network Configuration Control Via Connectivity Grah Processes Abubakr Muhammad Deartment of Electrical and Systems Engineering University of Pennsylvania Philadelhia, PA 90 abubakr@seas.uenn.edu Magnus
More informationFigure : An 8 bridge design grid. (a) Run this model using LOQO. What is the otimal comliance? What is the running time?
5.094/SMA53 Systems Otimization: Models and Comutation Assignment 5 (00 o i n ts) Due Aril 7, 004 Some Convex Analysis (0 o i n ts) (a) Given ositive scalars L and E, consider the following set in three-dimensional
More informationLinear diophantine equations for discrete tomography
Journal of X-Ray Science and Technology 10 001 59 66 59 IOS Press Linear diohantine euations for discrete tomograhy Yangbo Ye a,gewang b and Jiehua Zhu a a Deartment of Mathematics, The University of Iowa,
More informationCOMMUNICATION BETWEEN SHAREHOLDERS 1
COMMUNICATION BTWN SHARHOLDRS 1 A B. O A : A D Lemma B.1. U to µ Z r 2 σ2 Z + σ2 X 2r ω 2 an additive constant that does not deend on a or θ, the agents ayoffs can be written as: 2r rθa ω2 + θ µ Y rcov
More informationSTABILITY ANALYSIS AND CONTROL OF STOCHASTIC DYNAMIC SYSTEMS USING POLYNOMIAL CHAOS. A Dissertation JAMES ROBERT FISHER
STABILITY ANALYSIS AND CONTROL OF STOCHASTIC DYNAMIC SYSTEMS USING POLYNOMIAL CHAOS A Dissertation by JAMES ROBERT FISHER Submitted to the Office of Graduate Studies of Texas A&M University in artial fulfillment
More informationCSC165H, Mathematical expression and reasoning for computer science week 12
CSC165H, Mathematical exression and reasoning for comuter science week 1 nd December 005 Gary Baumgartner and Danny Hea hea@cs.toronto.edu SF4306A 416-978-5899 htt//www.cs.toronto.edu/~hea/165/s005/index.shtml
More informationMATH 2710: NOTES FOR ANALYSIS
MATH 270: NOTES FOR ANALYSIS The main ideas we will learn from analysis center around the idea of a limit. Limits occurs in several settings. We will start with finite limits of sequences, then cover infinite
More informationSolution of time domain electric field integral equation for arbitrarily shaped dielectric bodies using an unconditionally stable methodology
RADIO SCIENCE, VOL. 38, NO. 3, 148, doi:1.129/22rs2759, 23 Solution of time domain electric field integral equation for arbitrarily shaed dielectric bodies using an unconditionally stable methodology Young-seek
More informationThe Value of Even Distribution for Temporal Resource Partitions
The Value of Even Distribution for Temoral Resource Partitions Yu Li, Albert M. K. Cheng Deartment of Comuter Science University of Houston Houston, TX, 7704, USA htt://www.cs.uh.edu Technical Reort Number
More informationBrownian Motion and Random Prime Factorization
Brownian Motion and Random Prime Factorization Kendrick Tang June 4, 202 Contents Introduction 2 2 Brownian Motion 2 2. Develoing Brownian Motion.................... 2 2.. Measure Saces and Borel Sigma-Algebras.........
More informationFind the equation of a plane perpendicular to the line x = 2t + 1, y = 3t + 4, z = t 1 and passing through the point (2, 1, 3).
CME 100 Midterm Solutions - Fall 004 1 CME 100 - Midterm Solutions - Fall 004 Problem 1 Find the equation of a lane erendicular to the line x = t + 1, y = 3t + 4, z = t 1 and assing through the oint (,
More informationDistributed Rule-Based Inference in the Presence of Redundant Information
istribution Statement : roved for ublic release; distribution is unlimited. istributed Rule-ased Inference in the Presence of Redundant Information June 8, 004 William J. Farrell III Lockheed Martin dvanced
More informationPositive decomposition of transfer functions with multiple poles
Positive decomosition of transfer functions with multile oles Béla Nagy 1, Máté Matolcsi 2, and Márta Szilvási 1 Deartment of Analysis, Technical University of Budaest (BME), H-1111, Budaest, Egry J. u.
More informationThe Thermo Economical Cost Minimization of Heat Exchangers
he hermo Economical ost Minimization of Heat Exchangers Dr Möylemez Deartment of Mechanical Engineering, niversity of Gaziante, 7310 sait@ganteedutr bstract- thermo economic otimization analysis is resented
More informationGradient Based Iterative Algorithms for Solving a Class of Matrix Equations
1216 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 50, NO. 8, AUGUST 2005 Exanding (A3)gives M 0 8 T +1 T +18 =( T +1 +1 ) F = I (A4) g = 0F 8 T +1 =( T +1 +1 ) f = T +1 +1 + T +18 F 8 T +1 =( T +1 +1 )
More informationA STUDY ON THE UTILIZATION OF COMPATIBILITY METRIC IN THE AHP: APPLYING TO SOFTWARE PROCESS ASSESSMENTS
ISAHP 2005, Honolulu, Hawaii, July 8-10, 2003 A SUDY ON HE UILIZAION OF COMPAIBILIY MERIC IN HE AHP: APPLYING O SOFWARE PROCESS ASSESSMENS Min-Suk Yoon Yosu National University San 96-1 Dundeok-dong Yeosu
More information