A Unified View on Multi-class Support Vector Classification Supplement

Size: px
Start display at page:

Download "A Unified View on Multi-class Support Vector Classification Supplement"

Transcription

1 Journal of Mahine Learning Researh??) Submitted 7/15; Published?/?? A Unified View on Multi-lass Support Vetor Classifiation Supplement Ürün Doğan Mirosoft Researh Tobias Glasmahers Institut für Neuroinformatik Ruhr-Universität Bohum, Germany Christian Igel Department of Computer Siene University of Copenhagen, Denmark udogan@mirosoft.om tobias.glasmahers@ini.rub.de igel@diku.dk Editor: Ingo Steinwart A. Aggregation Operators As Linear Programs Aggregation operators, whih ombine the d margin violations into a single ost value, an be understood as omputing the value of the linear program v, y) = min ξ r R y ξ r s.t. p P y : ξ syp) v p fx), y) for the variables ξ = ξ r ) r Ry, where P y Y, R y is an index set, and s y : P y R y is surjetive. The set P y lists all margin violations that enter the loss, R y lists the slak variables, and s y assigns slak variables to margin omponents, depending on the partiular loss in use. Table 1 lists the onfigurations of the linear programs orresponding to the different aggregation operators. linear program definition aggregation operator P y R y s y self {y} { } p o-max Y \ {y} { } p t-max Y { } p t-sum Y Y id o-sum Y \ {y} Y \ {y} id Table 1: Aggregation operators and the orresponding linear programs, expressed in terms of the sets P y and R y, and the assignment s y : P y R y, for eah y Y. As for the margin funtion definition based on the sparse oeffiients ν y,p,m, the true underlying degrees of freedom for aggregation operators are far more restrited than it? Ürün Doğan, Tobias Glasmahers and Christian Igel.

2 Doğan, Glasmahers and Igel seems, in partiular if lasses are treated symmetrially. For symmetry reasons, the sets P y an take only the three values {y}, Y, and Y \ {y}, sine all lasses y are to be treated the same way. The same argument implies that s y either has to be injetive or onstant, restrited to the atomi invariant subsets {y} and Y \ {y}. This again leaves only few hoies for R y under the restrition that s y is surjetive. The hinge loss L hinge µ) = max{0, 1 µ} an also be expressed as a linear program, namely L hinge µ) = min u u s.t. u 1 µ u 0. The two linear programs an be ombined into one: Lfx), y) = min ξ r R y ξ r s.t. p P y : ξ syp) 1 µ p fx), y) r R y : ξ r 0 The first onstraint an be rewritten as µ p fx), y) = m ν y,p,m f m x) 1 ξ syp). Thus, the deision funtion values enter a multi-lass loss based on the hinge loss as parameters of a linear program. B. Deriving the Uniform Dual Problems For deriving the dual problem from the primal, we introdue Lagrange multipliers α 0, β i,r 0, η H, and τ R orresponding to the onstraints of the primal problem, and ompute the Lagrangian L = 1 w 2 + C ξ i,r 2 i,r + [ α γ yi,p ] ) ν yi,p, w, φx i ) + b ξi,syi p) β i,r ξ i,r i,r + η, w + τ b, 2

3 A Unified View on Multi-ategory Support Vetor Classifiation: Supplement with derivatives L = w α ν yi,p,φx i ) + η = 0 w = α ν yi,p,φx i ) η 1) w L = α ν yi,p, + τ = 0 α ν yi,p, = τ b L = C α β i,r = 0 α C. ξ i,r p P r y The sets Py r, r R y, are defined as Py r = s 1 y {r}) = {p P y s y p) = r}. They form a partition of the set P y of onstraints. To derive the dual in the absene of the sum-to-zero onstraint we just set the dual variables η and τ to zero. Then the first derivative above gives us an expression of w in terms of α. In the ase with sum-to-zero onstraint we get 0 = w = p P r y α ν yi,p,φx i ) η η = α 1 d ν yi,p, ) φx i ) and thus w = α [ m δ m, 1 ) ] ν yi,p,m φx i ). d To get to the dual problem, we plug this expression into the Lagrangian using the identity δ m, 1 ) δ n, 1 ) = δ m,n 1 ). d d d C. Proof of Theorem 5 In the following, we outline a proof of Theorem 5. Let Lfx), y) denote either the loss funtion used by the AMO mahine or the loss funtion used by the ATM mahine, that is, the loss resulting from appliation of either the max-over-others or the total-max operator to absolute margins: { } Lfx), y) = max v abs fx), y) = [ { 1 + max f x) }] AMO) Y \{y} Y \{y} + or { } Lfx), y) = max v abs fx), y) Y { [ ] = max 1 + max {f x)}, [ 1 f y x) ] } Y \{y} + + ATM) Then Theorem 5 states that the minimizer f of the orresponding risk R = E[Lfx), y)], subjet to the sum-to-zero onstraint Y f x) = 0, satisfies:. 3

4 Doğan, Glasmahers and Igel If there exists a majority lass y Y suh that P y > d 1)/d, then f y x) = d 1 and f x) = 1 for all Y \ {y}. If P y < d 1)/d for all y Y, then f x) = 0. Proof We demonstrate the proof for the AMO loss funtion. Following Liu 2007), we argue that f x) 1 for all Y. Suppose f x) < 1, then it is easy to see that f defined as f x) = 1 and f e x) = f e x) + f x) + 1)/d 1) fulfills R x f) R x f) ontraditing the optimality of f. Restriting the solution spae to f x) 1 allows us to write the point-wise risk as R x = y Y P y 1 + max { f x) } ) Y \ {y}. Now we pik y arg max{f x) Y } and treat the value f y x) 0 whih is non-negative beause of the sum-to-zero onstraint) as fixed from now on. We write the point-wise risk as R x = P y 1 + max { f x) } ) Y \ {y} + P 1 + f y x) ). y The best we an do to keep this risk low is to set all omponents f x), y, to the same value: f x) = y f x)/d 1) = f y x)/d 1) for all Y \ {y}. It holds R x = P y 1 f ) y) + P 1 + f y x) ) = P y 1 f y) d 1 d 1 y = 1 P y fy) d P y) f y x) = d d 1 P y ) f y x). ) + 1 P y ) 1 + f y x) ) For P y > d 1)/d this expression is a dereasing funtion of f y x), resulting in the optimum f y x) = d 1 and f x) = 1 for y, whih maximizes f y x) under the onstraints x f x) = 0 and : f x) 1. In ontrast, for P y < d 1)/d the risk is lower bounded by one. In this ase f x) = 0 minimizes the expression yielding R x = 1. The analogous result for the ATM loss funtion an be proven with exatly the same arguments. 4

5 A Unified View on Multi-ategory Support Vetor Classifiation: Supplement D. Data Sets The desriptive statistis of the 12 UCI data sets used in both the linear as well as nonlinear SVM experiments are given in Table 2. The additional data sets used in the linear SVM experiments are desribed in Table 3. Data set d l train l test p Abalone Car Glass Iris Opt. digits Page bloks Sat Segment Soy bean Vehile Red wine White wine Table 2: Desriptive statistis of the 12 UCI data sets used in the non-linear SVM study. The olumns d, l train, l test, and p ontain the number of lasses, the number of training examples, the number of test examples, and the input spae dimension number of features), respetively. Data set d l train l test p Covertype 7 406, , Letter 26 15,000 5, News ,935 3,993 62,061 Setor 105 6,412 3,207 55,197 Usps 10 7,291 2, Table 3: Desriptive statistis of the additional data sets used in the linear SVM experiments. The olumns d, l train, l test, and p ontain the number of lasses, the number of training examples, the number of test examples, and the input spae dimension number of features), respetively. 5

6 Doğan, Glasmahers and Igel E. Model Seletion Results The best parameter onfigurations C, γ) for the non-linear SVMs are found in Table 4. The values of the parameter C for the linear SVM experiments are listed in Table 5. OVA MMR WW CS LLW AMO ATS ATM RM Abalone Car Glass Iris Opt. digits Page bloks Sat Segment Soy bean Vehile Red wine White wine Table 4: Best hyperparameter values, ) found by the model seletion proedure. Referenes Y. Liu. Fisher onsisteny of multiategory support vetor mahines. In M. Meila and X. Shen, editors, Eleventh International Conferene on Artifiial Intelligene and Statistis AISTATS), volume 2 of JMLR W&P, pages ,

7 A Unified View on Multi-ategory Support Vetor Classifiation: Supplement OVA MMR WW CS LLW AMO ATS ATM RM Cover type Letter News Setor Usps Abalone Car Glass Iris Opt. digits Page bloks Sat Segment Soybean Vehile Red wine White wine Table 5: Best hyperparameter values ) for linear models found by the model seletion proedure. 7

Maximum Entropy and Exponential Families

Maximum Entropy and Exponential Families Maximum Entropy and Exponential Families April 9, 209 Abstrat The goal of this note is to derive the exponential form of probability distribution from more basi onsiderations, in partiular Entropy. It

More information

Complexity of Regularization RBF Networks

Complexity of Regularization RBF Networks Complexity of Regularization RBF Networks Mark A Kon Department of Mathematis and Statistis Boston University Boston, MA 02215 mkon@buedu Leszek Plaskota Institute of Applied Mathematis University of Warsaw

More information

Model-based mixture discriminant analysis an experimental study

Model-based mixture discriminant analysis an experimental study Model-based mixture disriminant analysis an experimental study Zohar Halbe and Mayer Aladjem Department of Eletrial and Computer Engineering, Ben-Gurion University of the Negev P.O.Box 653, Beer-Sheva,

More information

max min z i i=1 x j k s.t. j=1 x j j:i T j

max min z i i=1 x j k s.t. j=1 x j j:i T j AM 221: Advaned Optimization Spring 2016 Prof. Yaron Singer Leture 22 April 18th 1 Overview In this leture, we will study the pipage rounding tehnique whih is a deterministi rounding proedure that an be

More information

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION LOGISIC REGRESSIO I DEPRESSIO CLASSIFICAIO J. Kual,. V. ran, M. Bareš KSE, FJFI, CVU v Praze PCP, CS, 3LF UK v Praze Abstrat Well nown logisti regression and the other binary response models an be used

More information

Danielle Maddix AA238 Final Project December 9, 2016

Danielle Maddix AA238 Final Project December 9, 2016 Struture and Parameter Learning in Bayesian Networks with Appliations to Prediting Breast Caner Tumor Malignany in a Lower Dimension Feature Spae Danielle Maddix AA238 Final Projet Deember 9, 2016 Abstrat

More information

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM NETWORK SIMPLEX LGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM Cen Çalışan, Utah Valley University, 800 W. University Parway, Orem, UT 84058, 801-863-6487, en.alisan@uvu.edu BSTRCT The minimum

More information

Math 220A - Fall 2002 Homework 8 Solutions

Math 220A - Fall 2002 Homework 8 Solutions Math A - Fall Homework 8 Solutions 1. Consider u tt u = x R 3, t > u(x, ) = φ(x) u t (x, ) = ψ(x). Suppose φ, ψ are supported in the annular region a < x < b. (a) Find the time T 1 > suh that u(x, t) is

More information

Bilinear Formulated Multiple Kernel Learning for Multi-class Classification Problem

Bilinear Formulated Multiple Kernel Learning for Multi-class Classification Problem Bilinear Formulated Multiple Kernel Learning for Multi-lass Classifiation Problem Takumi Kobayashi and Nobuyuki Otsu National Institute of Advaned Industrial Siene and Tehnology, -- Umezono, Tsukuba, Japan

More information

Chapter 8 Hypothesis Testing

Chapter 8 Hypothesis Testing Leture 5 for BST 63: Statistial Theory II Kui Zhang, Spring Chapter 8 Hypothesis Testing Setion 8 Introdution Definition 8 A hypothesis is a statement about a population parameter Definition 8 The two

More information

Hankel Optimal Model Order Reduction 1

Hankel Optimal Model Order Reduction 1 Massahusetts Institute of Tehnology Department of Eletrial Engineering and Computer Siene 6.245: MULTIVARIABLE CONTROL SYSTEMS by A. Megretski Hankel Optimal Model Order Redution 1 This leture overs both

More information

A NONLILEAR CONTROLLER FOR SHIP AUTOPILOTS

A NONLILEAR CONTROLLER FOR SHIP AUTOPILOTS Vietnam Journal of Mehanis, VAST, Vol. 4, No. (), pp. A NONLILEAR CONTROLLER FOR SHIP AUTOPILOTS Le Thanh Tung Hanoi University of Siene and Tehnology, Vietnam Abstrat. Conventional ship autopilots are

More information

Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers)

Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers) Support vector machines In a nutshell Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers) Solution only depends on a small subset of training

More information

The Influences of Smooth Approximation Functions for SPTSVM

The Influences of Smooth Approximation Functions for SPTSVM The Influenes of Smooth Approximation Funtions for SPTSVM Xinxin Zhang Liaoheng University Shool of Mathematis Sienes Liaoheng, 5059 P.R. China ldzhangxin008@6.om Liya Fan Liaoheng University Shool of

More information

Lecture 7: Sampling/Projections for Least-squares Approximation, Cont. 7 Sampling/Projections for Least-squares Approximation, Cont.

Lecture 7: Sampling/Projections for Least-squares Approximation, Cont. 7 Sampling/Projections for Least-squares Approximation, Cont. Stat60/CS94: Randomized Algorithms for Matries and Data Leture 7-09/5/013 Leture 7: Sampling/Projetions for Least-squares Approximation, Cont. Leturer: Mihael Mahoney Sribe: Mihael Mahoney Warning: these

More information

Solutions to Problem Set 1

Solutions to Problem Set 1 Eon602: Maro Theory Eonomis, HKU Instrutor: Dr. Yulei Luo September 208 Solutions to Problem Set. [0 points] Consider the following lifetime optimal onsumption-saving problem: v (a 0 ) max f;a t+ g t t

More information

Research Article Approximation of Analytic Functions by Solutions of Cauchy-Euler Equation

Research Article Approximation of Analytic Functions by Solutions of Cauchy-Euler Equation Funtion Spaes Volume 2016, Artile ID 7874061, 5 pages http://d.doi.org/10.1155/2016/7874061 Researh Artile Approimation of Analyti Funtions by Solutions of Cauhy-Euler Equation Soon-Mo Jung Mathematis

More information

Weighted K-Nearest Neighbor Revisited

Weighted K-Nearest Neighbor Revisited Weighted -Nearest Neighbor Revisited M. Biego University of Verona Verona, Italy Email: manuele.biego@univr.it M. Loog Delft University of Tehnology Delft, The Netherlands Email: m.loog@tudelft.nl Abstrat

More information

QUANTUM MECHANICS II PHYS 517. Solutions to Problem Set # 1

QUANTUM MECHANICS II PHYS 517. Solutions to Problem Set # 1 QUANTUM MECHANICS II PHYS 57 Solutions to Problem Set #. The hamiltonian for a lassial harmoni osillator an be written in many different forms, suh as use ω = k/m H = p m + kx H = P + Q hω a. Find a anonial

More information

A Note on Extending Generalization Bounds for Binary Large-Margin Classifiers to Multiple Classes

A Note on Extending Generalization Bounds for Binary Large-Margin Classifiers to Multiple Classes A Note on Extending Generalization Bounds for Binary Large-Margin Classifiers to Multiple Classes Ürün Dogan 1 Tobias Glasmachers 2 and Christian Igel 3 1 Institut für Mathematik Universität Potsdam Germany

More information

Symplectic Projector and Physical Degrees of Freedom of The Classical Particle

Symplectic Projector and Physical Degrees of Freedom of The Classical Particle Sympleti Projetor and Physial Degrees of Freedom of The Classial Partile M. A. De Andrade a, M. A. Santos b and I. V. Vanea arxiv:hep-th/0308169v3 7 Sep 2003 a Grupo de Físia Teória, Universidade Católia

More information

An Integer Solution of Fractional Programming Problem

An Integer Solution of Fractional Programming Problem Gen. Math. Notes, Vol. 4, No., June 0, pp. -9 ISSN 9-784; Copyright ICSRS Publiation, 0 www.i-srs.org Available free online at http://www.geman.in An Integer Solution of Frational Programming Problem S.C.

More information

HYPERSTABILITY OF THE GENERAL LINEAR FUNCTIONAL EQUATION

HYPERSTABILITY OF THE GENERAL LINEAR FUNCTIONAL EQUATION Bull. Korean Math. So. 52 (2015, No. 6, pp. 1827 1838 http://dx.doi.org/10.4134/bkms.2015.52.6.1827 HYPERSTABILITY OF THE GENERAL LINEAR FUNCTIONAL EQUATION Magdalena Piszzek Abstrat. We give some results

More information

Tutorial 4 (week 4) Solutions

Tutorial 4 (week 4) Solutions THE UNIVERSITY OF SYDNEY PURE MATHEMATICS Summer Shool Math26 28 Tutorial week s You are given the following data points: x 2 y 2 Construt a Lagrange basis {p p p 2 p 3 } of P 3 using the x values from

More information

Stochastic Combinatorial Optimization with Risk Evdokia Nikolova

Stochastic Combinatorial Optimization with Risk Evdokia Nikolova Computer Siene and Artifiial Intelligene Laboratory Tehnial Report MIT-CSAIL-TR-2008-055 September 13, 2008 Stohasti Combinatorial Optimization with Risk Evdokia Nikolova massahusetts institute of tehnology,

More information

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1 Computer Siene 786S - Statistial Methods in Natural Language Proessing and Data Analysis Page 1 Hypothesis Testing A statistial hypothesis is a statement about the nature of the distribution of a random

More information

Nonreversibility of Multiple Unicast Networks

Nonreversibility of Multiple Unicast Networks Nonreversibility of Multiple Uniast Networks Randall Dougherty and Kenneth Zeger September 27, 2005 Abstrat We prove that for any finite direted ayli network, there exists a orresponding multiple uniast

More information

Advanced Computational Fluid Dynamics AA215A Lecture 4

Advanced Computational Fluid Dynamics AA215A Lecture 4 Advaned Computational Fluid Dynamis AA5A Leture 4 Antony Jameson Winter Quarter,, Stanford, CA Abstrat Leture 4 overs analysis of the equations of gas dynamis Contents Analysis of the equations of gas

More information

Differential Equations 8/24/2010

Differential Equations 8/24/2010 Differential Equations A Differential i Equation (DE) is an equation ontaining one or more derivatives of an unknown dependant d variable with respet to (wrt) one or more independent variables. Solution

More information

Chapter 2. Conditional Probability

Chapter 2. Conditional Probability Chapter. Conditional Probability The probabilities assigned to various events depend on what is known about the experimental situation when the assignment is made. For a partiular event A, we have used

More information

Sensitivity Analysis in Markov Networks

Sensitivity Analysis in Markov Networks Sensitivity Analysis in Markov Networks Hei Chan and Adnan Darwihe Computer Siene Department University of California, Los Angeles Los Angeles, CA 90095 {hei,darwihe}@s.ula.edu Abstrat This paper explores

More information

Normative and descriptive approaches to multiattribute decision making

Normative and descriptive approaches to multiattribute decision making De. 009, Volume 8, No. (Serial No.78) China-USA Business Review, ISSN 57-54, USA Normative and desriptive approahes to multiattribute deision making Milan Terek (Department of Statistis, University of

More information

7 Max-Flow Problems. Business Computing and Operations Research 608

7 Max-Flow Problems. Business Computing and Operations Research 608 7 Max-Flow Problems Business Computing and Operations Researh 68 7. Max-Flow Problems In what follows, we onsider a somewhat modified problem onstellation Instead of osts of transmission, vetor now indiates

More information

An I-Vector Backend for Speaker Verification

An I-Vector Backend for Speaker Verification An I-Vetor Bakend for Speaker Verifiation Patrik Kenny, 1 Themos Stafylakis, 1 Jahangir Alam, 1 and Marel Kokmann 2 1 CRIM, Canada, {patrik.kenny, themos.stafylakis, jahangir.alam}@rim.a 2 VoieTrust, Canada,

More information

Design and Development of Three Stages Mixed Sampling Plans for Variable Attribute Variable Quality Characteristics

Design and Development of Three Stages Mixed Sampling Plans for Variable Attribute Variable Quality Characteristics International Journal of Statistis and Systems ISSN 0973-2675 Volume 12, Number 4 (2017), pp. 763-772 Researh India Publiations http://www.ripubliation.om Design and Development of Three Stages Mixed Sampling

More information

SURFACE WAVES OF NON-RAYLEIGH TYPE

SURFACE WAVES OF NON-RAYLEIGH TYPE SURFACE WAVES OF NON-RAYLEIGH TYPE by SERGEY V. KUZNETSOV Institute for Problems in Mehanis Prosp. Vernadskogo, 0, Mosow, 75 Russia e-mail: sv@kuznetsov.msk.ru Abstrat. Existene of surfae waves of non-rayleigh

More information

Scalable system level synthesis for virtually localizable systems

Scalable system level synthesis for virtually localizable systems Salable system level synthesis for virtually loalizable systems Nikolai Matni, Yuh-Shyang Wang and James Anderson Abstrat In previous work, we developed the system level approah to ontroller synthesis,

More information

Københavns Universitet. Fast training of multi-class support vector machines Dogan, Ürün; Glasmachers, Tobias; Igel, Christian. Publication date: 2011

Københavns Universitet. Fast training of multi-class support vector machines Dogan, Ürün; Glasmachers, Tobias; Igel, Christian. Publication date: 2011 university of copenhagen Københavns Universitet Fast training of multi-class support vector machines Dogan, Ürün; Glasmachers, Tobias; Igel, Christian Publication date: 2011 Document Version Publisher's

More information

Training Multi-class Support Vector Machines. Dissertation

Training Multi-class Support Vector Machines. Dissertation Training Multi-class Support Vector Machines Dissertation zur Erlangung des Grades eines Doktor-Ingenieurs der Fakultät für Elektrotechnik und Informationstechnik an der Ruhr-Universität vorgelegt von

More information

The Hanging Chain. John McCuan. January 19, 2006

The Hanging Chain. John McCuan. January 19, 2006 The Hanging Chain John MCuan January 19, 2006 1 Introdution We onsider a hain of length L attahed to two points (a, u a and (b, u b in the plane. It is assumed that the hain hangs in the plane under a

More information

Methods of evaluating tests

Methods of evaluating tests Methods of evaluating tests Let X,, 1 Xn be i.i.d. Bernoulli( p ). Then 5 j= 1 j ( 5, ) T = X Binomial p. We test 1 H : p vs. 1 1 H : p>. We saw that a LRT is 1 if t k* φ ( x ) =. otherwise (t is the observed

More information

Sensitivity analysis for linear optimization problem with fuzzy data in the objective function

Sensitivity analysis for linear optimization problem with fuzzy data in the objective function Sensitivity analysis for linear optimization problem with fuzzy data in the objetive funtion Stephan Dempe, Tatiana Starostina May 5, 2004 Abstrat Linear programming problems with fuzzy oeffiients in the

More information

Modeling of discrete/continuous optimization problems: characterization and formulation of disjunctions and their relaxations

Modeling of discrete/continuous optimization problems: characterization and formulation of disjunctions and their relaxations Computers and Chemial Engineering (00) 4/448 www.elsevier.om/loate/omphemeng Modeling of disrete/ontinuous optimization problems: haraterization and formulation of disjuntions and their relaxations Aldo

More information

Probabilistic Graphical Models

Probabilistic Graphical Models Probabilisti Graphial Models David Sontag New York University Leture 12, April 19, 2012 Aknowledgement: Partially based on slides by Eri Xing at CMU and Andrew MCallum at UMass Amherst David Sontag (NYU)

More information

University of Groningen

University of Groningen University of Groningen Port Hamiltonian Formulation of Infinite Dimensional Systems II. Boundary Control by Interonnetion Mahelli, Alessandro; van der Shaft, Abraham; Melhiorri, Claudio Published in:

More information

An iterative least-square method suitable for solving large sparse matrices

An iterative least-square method suitable for solving large sparse matrices An iteratie least-square method suitable for soling large sparse matries By I. M. Khabaza The purpose of this paper is to report on the results of numerial experiments with an iteratie least-square method

More information

Product Policy in Markets with Word-of-Mouth Communication. Technical Appendix

Product Policy in Markets with Word-of-Mouth Communication. Technical Appendix rodut oliy in Markets with Word-of-Mouth Communiation Tehnial Appendix August 05 Miro-Model for Inreasing Awareness In the paper, we make the assumption that awareness is inreasing in ustomer type. I.e.,

More information

4.3 Singular Value Decomposition and Analysis

4.3 Singular Value Decomposition and Analysis 4.3 Singular Value Deomposition and Analysis A. Purpose Any M N matrix, A, has a Singular Value Deomposition (SVD) of the form A = USV t where U is an M M orthogonal matrix, V is an N N orthogonal matrix,

More information

CMSC 451: Lecture 9 Greedy Approximation: Set Cover Thursday, Sep 28, 2017

CMSC 451: Lecture 9 Greedy Approximation: Set Cover Thursday, Sep 28, 2017 CMSC 451: Leture 9 Greedy Approximation: Set Cover Thursday, Sep 28, 2017 Reading: Chapt 11 of KT and Set 54 of DPV Set Cover: An important lass of optimization problems involves overing a ertain domain,

More information

Journal of Inequalities in Pure and Applied Mathematics

Journal of Inequalities in Pure and Applied Mathematics Journal of Inequalities in Pure and Applied Mathematis A NEW ARRANGEMENT INEQUALITY MOHAMMAD JAVAHERI University of Oregon Department of Mathematis Fenton Hall, Eugene, OR 97403. EMail: javaheri@uoregon.edu

More information

EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION

EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION Journal of Mathematial Sienes: Advanes and Appliations Volume 3, 05, Pages -3 EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION JIAN YANG, XIAOJUAN LU and SHENGQIANG TANG

More information

Taste for variety and optimum product diversity in an open economy

Taste for variety and optimum product diversity in an open economy Taste for variety and optimum produt diversity in an open eonomy Javier Coto-Martínez City University Paul Levine University of Surrey Otober 0, 005 María D.C. Garía-Alonso University of Kent Abstrat We

More information

A Queueing Model for Call Blending in Call Centers

A Queueing Model for Call Blending in Call Centers A Queueing Model for Call Blending in Call Centers Sandjai Bhulai and Ger Koole Vrije Universiteit Amsterdam Faulty of Sienes De Boelelaan 1081a 1081 HV Amsterdam The Netherlands E-mail: {sbhulai, koole}@s.vu.nl

More information

23.1 Tuning controllers, in the large view Quoting from Section 16.7:

23.1 Tuning controllers, in the large view Quoting from Section 16.7: Lesson 23. Tuning a real ontroller - modeling, proess identifiation, fine tuning 23.0 Context We have learned to view proesses as dynami systems, taking are to identify their input, intermediate, and output

More information

Estimating the probability law of the codelength as a function of the approximation error in image compression

Estimating the probability law of the codelength as a function of the approximation error in image compression Estimating the probability law of the odelength as a funtion of the approximation error in image ompression François Malgouyres Marh 7, 2007 Abstrat After some reolletions on ompression of images using

More information

Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers)

Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers) Support vector machines In a nutshell Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers) Solution only depends on a small subset of training

More information

Understanding Elementary Landscapes

Understanding Elementary Landscapes Understanding Elementary Landsapes L. Darrell Whitley Andrew M. Sutton Adele E. Howe Department of Computer Siene Colorado State University Fort Collins, CO 853 {whitley,sutton,howe}@s.olostate.edu ABSTRACT

More information

The gravitational phenomena without the curved spacetime

The gravitational phenomena without the curved spacetime The gravitational phenomena without the urved spaetime Mirosław J. Kubiak Abstrat: In this paper was presented a desription of the gravitational phenomena in the new medium, different than the urved spaetime,

More information

ELECTROMAGNETIC WAVES WITH NONLINEAR DISPERSION LAW. P. М. Меdnis

ELECTROMAGNETIC WAVES WITH NONLINEAR DISPERSION LAW. P. М. Меdnis ELECTROMAGNETIC WAVES WITH NONLINEAR DISPERSION LAW P. М. Меdnis Novosibirs State Pedagogial University, Chair of the General and Theoretial Physis, Russia, 636, Novosibirs,Viljujsy, 8 e-mail: pmednis@inbox.ru

More information

MODELING MATTER AT NANOSCALES. 4. Introduction to quantum treatments Eigenvectors and eigenvalues of a matrix

MODELING MATTER AT NANOSCALES. 4. Introduction to quantum treatments Eigenvectors and eigenvalues of a matrix MODELING MATTER AT NANOSCALES 4 Introdution to quantum treatments 403 Eigenvetors and eigenvalues of a matrix Simultaneous equations in the variational method The problem of simultaneous equations in the

More information

Geometry of Transformations of Random Variables

Geometry of Transformations of Random Variables Geometry of Transformations of Random Variables Univariate distributions We are interested in the problem of finding the distribution of Y = h(x) when the transformation h is one-to-one so that there is

More information

Q2. [40 points] Bishop-Hill Model: Calculation of Taylor Factors for Multiple Slip

Q2. [40 points] Bishop-Hill Model: Calculation of Taylor Factors for Multiple Slip 27-750, A.D. Rollett Due: 20 th Ot., 2011. Homework 5, Volume Frations, Single and Multiple Slip Crystal Plastiity Note the 2 extra redit questions (at the end). Q1. [40 points] Single Slip: Calulating

More information

ON LOWER LIPSCHITZ CONTINUITY OF MINIMAL POINTS. Ewa M. Bednarczuk

ON LOWER LIPSCHITZ CONTINUITY OF MINIMAL POINTS. Ewa M. Bednarczuk Disussiones Mathematiae Differential Inlusions, Control and Optimization 20 2000 ) 245 255 ON LOWER LIPSCHITZ CONTINUITY OF MINIMAL POINTS Ewa M. Bednarzuk Systems Researh Institute, PAS 01 447 Warsaw,

More information

Assessing the Performance of a BCI: A Task-Oriented Approach

Assessing the Performance of a BCI: A Task-Oriented Approach Assessing the Performane of a BCI: A Task-Oriented Approah B. Dal Seno, L. Mainardi 2, M. Matteui Department of Eletronis and Information, IIT-Unit, Politenio di Milano, Italy 2 Department of Bioengineering,

More information

3.2 Gaussian (Normal) Random Numbers and Vectors

3.2 Gaussian (Normal) Random Numbers and Vectors 3.2 Gaussian (Normal) Random Numbers and Vetors A. Purpose Generate pseudorandom numbers or vetors from the Gaussian (normal) distribution. B. Usage B.1 Generating Gaussian (normal) pseudorandom numbers

More information

On the Complexity of the Weighted Fused Lasso

On the Complexity of the Weighted Fused Lasso ON THE COMPLEXITY OF THE WEIGHTED FUSED LASSO On the Compleity of the Weighted Fused Lasso José Bento jose.bento@b.edu Ralph Furmaniak rf@am.org Surjyendu Ray rays@b.edu Abstrat The solution path of the

More information

Robust Flight Control Design for a Turn Coordination System with Parameter Uncertainties

Robust Flight Control Design for a Turn Coordination System with Parameter Uncertainties Amerian Journal of Applied Sienes 4 (7): 496-501, 007 ISSN 1546-939 007 Siene Publiations Robust Flight ontrol Design for a urn oordination System with Parameter Unertainties 1 Ari Legowo and Hiroshi Okubo

More information

Support Vector Machines, Kernel SVM

Support Vector Machines, Kernel SVM Support Vector Machines, Kernel SVM Professor Ameet Talwalkar Professor Ameet Talwalkar CS260 Machine Learning Algorithms February 27, 2017 1 / 40 Outline 1 Administration 2 Review of last lecture 3 SVM

More information

the following action R of T on T n+1 : for each θ T, R θ : T n+1 T n+1 is defined by stated, we assume that all the curves in this paper are defined

the following action R of T on T n+1 : for each θ T, R θ : T n+1 T n+1 is defined by stated, we assume that all the curves in this paper are defined How should a snake turn on ie: A ase study of the asymptoti isoholonomi problem Jianghai Hu, Slobodan N. Simić, and Shankar Sastry Department of Eletrial Engineering and Computer Sienes University of California

More information

Generalized Dimensional Analysis

Generalized Dimensional Analysis #HUTP-92/A036 7/92 Generalized Dimensional Analysis arxiv:hep-ph/9207278v1 31 Jul 1992 Howard Georgi Lyman Laboratory of Physis Harvard University Cambridge, MA 02138 Abstrat I desribe a version of so-alled

More information

Grasp Planning: How to Choose a Suitable Task Wrench Space

Grasp Planning: How to Choose a Suitable Task Wrench Space Grasp Planning: How to Choose a Suitable Task Wrenh Spae Ch. Borst, M. Fisher and G. Hirzinger German Aerospae Center - DLR Institute for Robotis and Mehatronis 8223 Wessling, Germany Email: [Christoph.Borst,

More information

Support Vector Machine (continued)

Support Vector Machine (continued) Support Vector Machine continued) Overlapping class distribution: In practice the class-conditional distributions may overlap, so that the training data points are no longer linearly separable. We need

More information

Ordered fields and the ultrafilter theorem

Ordered fields and the ultrafilter theorem F U N D A M E N T A MATHEMATICAE 59 (999) Ordered fields and the ultrafilter theorem by R. B e r r (Dortmund), F. D e l o n (Paris) and J. S h m i d (Dortmund) Abstrat. We prove that on the basis of ZF

More information

Factorized Asymptotic Bayesian Inference for Mixture Modeling

Factorized Asymptotic Bayesian Inference for Mixture Modeling Fatorized Asymptoti Bayesian Inferene for Mixture Modeling Ryohei Fujimaki NEC Laboratories Ameria Department of Media Analytis rfujimaki@sv.ne-labs.om Satoshi Morinaga NEC Corporation Information and

More information

Moments and Wavelets in Signal Estimation

Moments and Wavelets in Signal Estimation Moments and Wavelets in Signal Estimation Edward J. Wegman 1 Center for Computational Statistis George Mason University Hung T. Le 2 International usiness Mahines Abstrat: The problem of generalized nonparametri

More information

An Integrated Architecture of Adaptive Neural Network Control for Dynamic Systems

An Integrated Architecture of Adaptive Neural Network Control for Dynamic Systems An Integrated Arhiteture of Adaptive Neural Network Control for Dynami Systems Robert L. Tokar 2 Brian D.MVey2 'Center for Nonlinear Studies, 2Applied Theoretial Physis Division Los Alamos National Laboratory,

More information

9 Geophysics and Radio-Astronomy: VLBI VeryLongBaseInterferometry

9 Geophysics and Radio-Astronomy: VLBI VeryLongBaseInterferometry 9 Geophysis and Radio-Astronomy: VLBI VeryLongBaseInterferometry VLBI is an interferometry tehnique used in radio astronomy, in whih two or more signals, oming from the same astronomial objet, are reeived

More information

Control Theory association of mathematics and engineering

Control Theory association of mathematics and engineering Control Theory assoiation of mathematis and engineering Wojieh Mitkowski Krzysztof Oprzedkiewiz Department of Automatis AGH Univ. of Siene & Tehnology, Craow, Poland, Abstrat In this paper a methodology

More information

Viewing the Rings of a Tree: Minimum Distortion Embeddings into Trees

Viewing the Rings of a Tree: Minimum Distortion Embeddings into Trees Viewing the Rings of a Tree: Minimum Distortion Embeddings into Trees Amir Nayyeri Benjamin Raihel Abstrat We desribe a 1+ε) approximation algorithm for finding the minimum distortion embedding of an n-point

More information

Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers)

Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers) Support vector machines In a nutshell Linear classifiers selecting hyperplane maximizing separation margin between classes (large margin classifiers) Solution only depends on a small subset of training

More information

Experiment 3: Basic Electronic Circuits II (tbc 1/7/2007)

Experiment 3: Basic Electronic Circuits II (tbc 1/7/2007) Experiment 3: Basi Eletroni iruits II (tb /7/007) Objetive: a) To study the first-order dynamis of a apaitive iruits with the appliation of Kirhoff s law, Ohm s law and apaitane formula. b) To learn how

More information

Transformation to approximate independence for locally stationary Gaussian processes

Transformation to approximate independence for locally stationary Gaussian processes ransformation to approximate independene for loally stationary Gaussian proesses Joseph Guinness, Mihael L. Stein We provide new approximations for the likelihood of a time series under the loally stationary

More information

A NEW FLEXIBLE BODY DYNAMIC FORMULATION FOR BEAM STRUCTURES UNDERGOING LARGE OVERALL MOTION IIE THREE-DIMENSIONAL CASE. W. J.

A NEW FLEXIBLE BODY DYNAMIC FORMULATION FOR BEAM STRUCTURES UNDERGOING LARGE OVERALL MOTION IIE THREE-DIMENSIONAL CASE. W. J. A NEW FLEXIBLE BODY DYNAMIC FORMULATION FOR BEAM STRUCTURES UNDERGOING LARGE OVERALL MOTION IIE THREE-DIMENSIONAL CASE W. J. Haering* Senior Projet Engineer General Motors Corporation Warren, Mihigan R.

More information

On the Licensing of Innovations under Strategic Delegation

On the Licensing of Innovations under Strategic Delegation On the Liensing of Innovations under Strategi Delegation Judy Hsu Institute of Finanial Management Nanhua University Taiwan and X. Henry Wang Department of Eonomis University of Missouri USA Abstrat This

More information

Integration of the Finite Toda Lattice with Complex-Valued Initial Data

Integration of the Finite Toda Lattice with Complex-Valued Initial Data Integration of the Finite Toda Lattie with Complex-Valued Initial Data Aydin Huseynov* and Gusein Sh Guseinov** *Institute of Mathematis and Mehanis, Azerbaijan National Aademy of Sienes, AZ4 Baku, Azerbaijan

More information

Counting Idempotent Relations

Counting Idempotent Relations Counting Idempotent Relations Beriht-Nr. 2008-15 Florian Kammüller ISSN 1436-9915 2 Abstrat This artile introdues and motivates idempotent relations. It summarizes haraterizations of idempotents and their

More information

SQUARE ROOTS AND AND DIRECTIONS

SQUARE ROOTS AND AND DIRECTIONS SQUARE ROOS AND AND DIRECIONS We onstrut a lattie-like point set in the Eulidean plane that eluidates the relationship between the loal statistis of the frational parts of n and diretions in a shifted

More information

Coefficients of the Inverse of Strongly Starlike Functions

Coefficients of the Inverse of Strongly Starlike Functions BULLETIN of the MALAYSIAN MATHEMATICAL SCIENCES SOCIETY Bull. Malaysian Math. S. So. (Seond Series) 6 (00) 6 7 Coeffiients of the Inverse of Strongly Starlie Funtions ROSIHAN M. ALI Shool of Mathematial

More information

11.1 Polynomial Least-Squares Curve Fit

11.1 Polynomial Least-Squares Curve Fit 11.1 Polynomial Least-Squares Curve Fit A. Purpose This subroutine determines a univariate polynomial that fits a given disrete set of data in the sense of minimizing the weighted sum of squares of residuals.

More information

The Corpuscular Structure of Matter, the Interaction of Material Particles, and Quantum Phenomena as a Consequence of Selfvariations.

The Corpuscular Structure of Matter, the Interaction of Material Particles, and Quantum Phenomena as a Consequence of Selfvariations. The Corpusular Struture of Matter, the Interation of Material Partiles, and Quantum Phenomena as a Consequene of Selfvariations. Emmanuil Manousos APM Institute for the Advanement of Physis and Mathematis,

More information

IDENTIFICATION AND CONTROL OF ACOUSTIC RADIATION MODES

IDENTIFICATION AND CONTROL OF ACOUSTIC RADIATION MODES IDENTIFICATION AND CONTROL OF ACOUSTIC RADIATION MODES Arthur P. Berkhoff University of Twente, Faulty of Eletrial Engineering, P.O. Box 217, 7 AE Enshede, The Netherlands email: a.p.berkhoff@el.utwente.nl

More information

I F I G R e s e a r c h R e p o r t. Minimal and Hyper-Minimal Biautomata. IFIG Research Report 1401 March Institut für Informatik

I F I G R e s e a r c h R e p o r t. Minimal and Hyper-Minimal Biautomata. IFIG Research Report 1401 March Institut für Informatik I F I G R e s e a r h R e p o r t Institut für Informatik Minimal and Hyper-Minimal Biautomata Markus Holzer Seastian Jakoi IFIG Researh Report 1401 Marh 2014 Institut für Informatik JLU Gießen Arndtstraße

More information

BAYES CLASSIFIER. Ivan Michael Siregar APLYSIT IT SOLUTION CENTER. Jl. Ir. H. Djuanda 109 Bandung

BAYES CLASSIFIER. Ivan Michael Siregar APLYSIT IT SOLUTION CENTER. Jl. Ir. H. Djuanda 109 Bandung BAYES CLASSIFIER www.aplysit.om www.ivan.siregar.biz ALYSIT IT SOLUTION CENTER Jl. Ir. H. Duanda 109 Bandung Ivan Mihael Siregar ivan.siregar@gmail.om Data Mining 2010 Bayesian Method Our fous this leture

More information

CSC2515 Winter 2015 Introduc3on to Machine Learning. Lecture 5: Clustering, mixture models, and EM

CSC2515 Winter 2015 Introduc3on to Machine Learning. Lecture 5: Clustering, mixture models, and EM CSC2515 Winter 2015 Introdu3on to Mahine Learning Leture 5: Clustering, mixture models, and EM All leture slides will be available as.pdf on the ourse website: http://www.s.toronto.edu/~urtasun/ourses/csc2515/

More information

Complete Shrimp Game Solution

Complete Shrimp Game Solution Complete Shrimp Game Solution Florian Ederer Feruary 7, 207 The inverse demand urve is given y P (Q a ; Q ; Q ) = 00 0:5 (Q a + Q + Q ) The pro t funtion for rm i = fa; ; g is i (Q a ; Q ; Q ) = Q i [P

More information

A note on a variational formulation of electrodynamics

A note on a variational formulation of electrodynamics Proeedings of the XV International Workshop on Geometry and Physis Puerto de la Cruz, Tenerife, Canary Islands, Spain September 11 16, 006 Publ. de la RSME, Vol. 11 (007), 314 31 A note on a variational

More information

Review of Force, Stress, and Strain Tensors

Review of Force, Stress, and Strain Tensors Review of Fore, Stress, and Strain Tensors.1 The Fore Vetor Fores an be grouped into two broad ategories: surfae fores and body fores. Surfae fores are those that at over a surfae (as the name implies),

More information

Chapter 9. The excitation process

Chapter 9. The excitation process Chapter 9 The exitation proess qualitative explanation of the formation of negative ion states Ne and He in He-Ne ollisions an be given by using a state orrelation diagram. state orrelation diagram is

More information

Sufficient Conditions for a Flexible Manufacturing System to be Deadlocked

Sufficient Conditions for a Flexible Manufacturing System to be Deadlocked Paper 0, INT 0 Suffiient Conditions for a Flexile Manufaturing System to e Deadloked Paul E Deering, PhD Department of Engineering Tehnology and Management Ohio University deering@ohioedu Astrat In reent

More information

Frugality Ratios And Improved Truthful Mechanisms for Vertex Cover

Frugality Ratios And Improved Truthful Mechanisms for Vertex Cover Frugality Ratios And Improved Truthful Mehanisms for Vertex Cover Edith Elkind Hebrew University of Jerusalem, Israel, and University of Southampton, Southampton, SO17 1BJ, U.K. Leslie Ann Goldberg University

More information