Algorismes IRWLS. May 2015

Size: px
Start display at page:

Download "Algorismes IRWLS. May 2015"

Transcription

1 Algorsmes IRWLS May 2015

2 1 Introducton Ths report studes the Iteratvely Re-Weghted Least Squares (IRWLS) algorthm for fndng the Maxmum Lkelhood (ML) estmators of a Generalsed Lnear Model (GLM) coeffcents. The data s related to metal treatments. Four dfferent temperatures z, {1,, 4} are appled to metal ngots, and the result s evaluated wth a bnary varable whch explans whether the ngot s ready for rollng or not. Let x be the number of successes (treatments not ready) and n the number of gnots tested n the condtons of temperature z n trals x notready z heat Table 1: Data for the study In the frst part of the practce the maxmum lkelhood estmators are found assumng that our data follows a Bnomal dstrbuton. X B(n, p (β)), = 1,, 4, β = (β 0, β 1 ), p (β) = eβ0+β z 1 + e β0+β z For ths purpose, the optm functon of R s used. Ths functon mplements solvers for unconstraned optmzaton problems. Wth the transformaton k (x) = x/(1 + x) we have a problem wthout restrctons on the parameters. Dfferent solvers are tested, ncludng the dervatve-free Nelder-Mead method. In the second part of the report a dfferent model s assumed. A GLM of the posson famly s consdered wth a logarthmc lnk functon. X P osson(λ ), = 1,, 4 log(λ ) = β 0 + β 1 n + β 2 z, = 1,, 4 The algorthm IRWLS s derved from the Fsher Scorng method and mplemented n R. Its results are compared wth the functon glm to prove that the same result s attaned. 2 Bnomal Dstrbuton Assumpton 2.1 Lkelhood Functon It s known that for one Bnomal dstrbuton the probablty densty functon s f x = n p x (1 p ) n x The log-lkelhood functon of the four samples wth the p defned above altogether s 4 L = log n + x log(p ) + (n x ) log(1 p ) x x We can plot the contour of ths functon as t s 2-Dmensonal wth the followng R-code 2

3 1 p r n t. l o g l l < functon ( Beta0, Beta1, x1=0,x2=2,x3=7,x4=3, z1 =7, z2 =14, z3 =27, z4 =57,n1=55,n2=157, 2 n3=159,n4=16){ 3 log ( ( f a c t o r a l ( n1 )/( f a c t o r a l ( x1 ) f a c t o r a l ( n1 x1 ))))+ x1 log ( ( exp( Beta0+Beta1 z1 )/(1 4 +exp( Beta0+Beta1 z1 ) ) ) ) + ( n1 x1 ) log (1 (exp( Beta0+Beta1 z1 )/(1+exp( Beta0+Beta1 z1 ) ) ) ) 5 +log ( ( f a c t o r a l ( n2 )/( f a c t o r a l ( x2 ) f a c t o r a l ( n2 x2 ))))+ x2 log ( ( exp( Beta0+Beta1 z2 )/(1 6 +exp( Beta0+Beta1 z2 ) ) ) ) + ( n2 x2 ) log (1 (exp( Beta0+Beta1 z2 )/(1+exp( Beta0+Beta1 z2 ) ) ) ) 7 +log ( ( f a c t o r a l ( n3 )/( f a c t o r a l ( x3 ) f a c t o r a l ( n3 x3 ))))+ x3 log ( ( exp( Beta0+Beta1 z3 )/(1 8 +exp( Beta0+Beta1 z3 ) ) ) ) + ( n3 x3 ) log (1 (exp( Beta0+Beta1 z3 )/(1+exp( Beta0+Beta1 z3 ) ) ) ) 9 +log ( ( f a c t o r a l ( n4 )/( f a c t o r a l ( x4 ) f a c t o r a l ( n4 x4 ))))+ x4 log ( ( exp( Beta0+Beta1 z4 )/(1 10 +exp( Beta0+Beta1 z4 ) ) ) ) + ( n4 x4 ) log (1 (exp( Beta0+Beta1 z4 )/(1+exp( Beta0+Beta1 z4 ) ) ) ) 11 } # Plot o f the f u n c t o n 14 pb0 < seq( 6, 4,length=101) 15 pb1 < seq (0.05, 0.09, length=101) 16 l o g l l. B 0. B 1 < outer (pb0, pb1, functon (B0, B1){ p r n t. l o g l l (B0, B1 ) } ) 17 wndows ( ) 18 contour (pb0, pb1, l o g l l. B 0. B 1, nlevels =100) Fgure 1: Lkelhood functon A maxmum s observed at a pont around 5 and In order to use the functon optm the decson varables must be defned as a numerc vector. Hence, a slght modfcaton must be ntroduced: 1 l o g l l < functon ( Beta0Beta1, x1=0,x2=2,x3=7,x4=3, z1 =7, z2 =14, z3 =27, z4 =57,n1=55,n2= Beta0 = Beta0Beta1 [ 1 ] 3 Beta1 = Beta0Beta1 [ 2 ] ( log ( ( f a c t o r a l ( n1 )/( f a c t o r a l ( x1 ) f a c t o r a l ( n1 x1 ))))+ x1 log ( ( exp( Beta } Notce that the return s multpled by mnus one. Ths way we face a mnmzaton problem. The same result can be acheved passng an argument to the functon argument. 3

4 2.2 Functon Scores In order to derve the gradent of the log-lkelhood equaton we know that L 4 ( ) β = p β (1 p ) β x + (1 p ) n (1) β p 1 p 1 p Hence, we can defne the followng 2 4 matrx Let us also defne the 2 1 matrx N And the gradent can be stated as Mj = 1 p j N = p j + 1 p j (2) β 1 1 p j β 1 4 k=1 1 p k n k (3) 1 p k β 1 L = M x + N, t(x) = [x 1 x 2 x 3 x 4 ] (4) 1 2 g r a d. l l < functon ( Beta0Beta1, x1=0,x2=2,x3=7,x4=3, z1 =7, z2 =14, z3 =27, z4 =57,n1=55, n Beta0 = Beta0Beta1 [ 1 ] 4 Beta1 = Beta0Beta1 [ 2 ] 5 6 p1 = exp( Beta0+Beta1 z1 )/(1+exp( Beta0+Beta1 z1 ) ) 7 p2 = exp( Beta0+Beta1 z2 )/(1+exp( Beta0+Beta1 z2 ) ) 8 p3 = exp( Beta0+Beta1 z3 )/(1+exp( Beta0+Beta1 z3 ) ) 9 p4 = exp( Beta0+Beta1 z4 )/(1+exp( Beta0+Beta1 z4 ) ) d f f. p 1. B 0 = exp( Beta1 z1+beta0 )/(1+exp( Beta0+Beta1 z1 )) exp(2 Beta1 z d f f. p 2. B 0 = exp( Beta1 z2+beta0 )/(1+exp( Beta0+Beta1 z2 )) exp(2 Beta1 z d f f. p 3. B 0 = exp( Beta1 z3+beta0 )/(1+exp( Beta0+Beta1 z3 )) exp(2 Beta1 z d f f. p 4. B 0 = exp( Beta1 z4+beta0 )/(1+exp( Beta0+Beta1 z4 )) exp(2 Beta1 z d f f. p 1. B 1 = ( z1 exp( Beta1 z1+beta0 ) ) /(exp( Beta1 z1+beta0 )+1) ( z1 exp(2 B e d f f. p 2. B 1 = ( z2 exp( Beta1 z2+beta0 ) ) /(exp( Beta1 z2+beta0 )+1) ( z2 exp(2 B e d f f. p 3. B 1 = ( z3 exp( Beta1 z3+beta0 ) ) /(exp( Beta1 z3+beta0 )+1) ( z3 exp(2 B e d f f. p 4. B 1 = ( z4 exp( Beta1 z4+beta0 ) ) /(exp( Beta1 z4+beta0 )+1) ( z4 exp(2 B e M11 = 1/p1 d f f. p 1. B 0 + 1/(1 p1 ) d f f. p 1. B 0 22 M12 = 1/p2 d f f. p 2. B 0 + 1/(1 p2 ) d f f. p 2. B 0 23 M13 = 1/p3 d f f. p 3. B 0 + 1/(1 p3 ) d f f. p 3. B 0 24 M14 = 1/p4 d f f. p 4. B 0 + 1/(1 p4 ) d f f. p 4. B M21 = 1/p1 d f f. p 1. B 1 + 1/(1 p1 ) d f f. p 1. B 1 27 M22 = 1/p2 d f f. p 2. B 1 + 1/(1 p2 ) d f f. p 2. B 1 28 M23 = 1/p3 d f f. p 3. B 1 + 1/(1 p3 ) d f f. p 3. B 1 29 M24 = 1/p4 d f f. p 4. B 1 + 1/(1 p4 ) d f f. p 4. B N1 = 1/(1 p1 ) d f f. p 1. B 0 n1 1/(1 p2 ) d f f. p 2. B 0 n2 1/(1 p3 ) d f f. p 3. B N2 = 1/(1 p1 ) d f f. p 1. B 1 n1 1/(1 p2 ) d f f. p 2. B 1 n2 1/(1 p3 ) d f f. p 3. B M = matrx ( c (M11, M12, M13, M14, M21, M22, M23, M24), byrow=t, nrow=2) 35 N = matrx ( c (N1, N2),nrow=2) X = matrx ( c ( x1, x2, x3, x4 ),nrow=4) return( 1 (M% %X+N) ) 40 } 4

5 2.3 Optmzaton of ML functon In ths secton dfferent optmzaton algorthms are consdered for mnmzng the mnus logarthm of the lkelhood functon. The default solver s the Nelder-Mead algorthm, whch does not consder gradents but only functon evaluatons. For the other algorthms, the gradent s entered as a parameter. If nformaton of the optmzaton procedure s desred, the nput parameter trace s to be used. Wth the Hessan parameter set to TRUE a numerc approxmaton at the optmum pont s gven. When t comes to the Conjugate Gradent, there s a specfc command for choosng a step rule: type. The maxmum number of teratons can be enlarged wth the parameter maxt. All ntal ponts are set to (0, 0). Notce that the Hessan needs not to be computed for these methods, and a cheaper approxmaton s used nstead. In the prevous exercses, where de Fsher Scorng algorthm was studed, the Hessan matrx was derved analytcally for some cases. The general purpose optmzaton algorthms assume the Hessan to be dramatcally more expensve to compute than the gradent or the objectve functon. However, n convex problems wth an easy-to-compute Hessan matrx other algorthms may be more effcent. 1 2 X < data.frame ( matrx ( rep ( 0, 2 0 ),nrow=5)) 3 names(x) < c ( Nelder Mead, BFGS, CG, L BFGS B ) 4 rownames(x) < c ( counts f o b j., counts grad, s o l. B 0, s o l. B 1, o b j. v a l ) # a. 1 ) Nelder Mead 8 opt1 < optm( par = c ( 0, 0 ), fn = l o g l l, control = l s t ( trace =1)) 9 # a. 2 ) BFGS 10 opt2 < optm( par = c ( 0, 0 ), fn = l o g l l, gr = g r a d. l l, method = BFGS, h e s s a n=t) 11 # a. 3 ) CG 12 opt3 < optm( par = c ( 0, 0 ), fn = l o g l l, gr = g r a d. l l, method = CG, 13 control = l s t ( maxt =10000, type =3), h e s s a n=t) 14 # a. 4 ) L BFGS B 15 opt4 < optm( par = c ( 0, 0 ), fn = l o g l l, gr = g r a d. l l, method = L BFGS B, h e s s a n=t) # S o l u t o n > X 21 Nelder Mead BFGS CG L BFGS B 22 counts f o b j e counts grad NA e s o l. B e s o l. B e o b j. v a l e All the algorthms attan a smlar soluton. The Conjugate Gradent (CG) s not able to satsfy ts nternal optmalty tolerance and reaches the maxmum number of teratons. The Nelder-Mead method needs to evaluate the objectve functon 81 tmes. the BFGS method reduces ths number to 46 at the cost of evaluatng the gradent 14 tmes. When the gradent s known, dervatve-free methods are not recommended, especally when functons are non-convex. The BFGS varant evaluates 22 tmes both the gradent and the objectve functon. The selecton of the method depends on the computatonal cost of evaluatng the objectve functon and the gradent. 5

6 2.4 Covarance matrx estmaton As stated before, a numercal approxmaton to the Hessan can be obtaned wth optm. The nverse of ths matrx s an approxmaton to the covarance matrx. 1 solve ( opt2$ h e s s a n ) 2 # S o l u t o n 3 4 > solve ( opt2$ h e s s a n ) 5 [, 1 ] [, 2 ] 6 [ 1, ] [ 2, ] Ths result can be compared to the result obtaned usng the Fshers Informaton computed analytcally. It s easy to compute the Fsher s Informaton matrx as the varance of the scores functon. V ar(m x + N) = M V ar(x) t(m) + 0 (5) 1 F s h e r. n f < functon ( Beta0, Beta1, x1=0,x2=2,x3=7,x4=3, z1 =7, z2 =14, z3 =27, z4= (... ) 4 v.x1 < n1 p1 (1 p1 ) 5 v.x2 < n2 p2 (1 p2 ) 6 v.x3 < n3 p3 (1 p3 ) 7 v.x4 < n4 p4 (1 p4 ) 8 9 Var.X < dag ( c ( v.x1, v.x2, v.x3, v.x4 ) ) return (M% %Var.X% %t (M) ) # S o l u t o n > solve ( F s h e r. n f ( , ) ) 16 [, 1 ] [, 2 ] 17 [ 1, ] [ 2, ] As t was expected the Hessan provded wth the functon optm s a very good approxmaton to the analtcally derved Fshers nformaton matrx. Hence, t can be used to estmate the varance of the maxmum lkelhood estmators. 2.5 Bootstrap Covarance matrx estmaton In ths secton a parametrc bootstrap procedure s used to estmate the varance of the estmators wth a dervatve-free method. Our Data s assumed to follow a bnomal dstrbuton wth the parameters obtaned maxmzng the lkelhood functon. In each teraton the maxmum lkelhood estmator must be obtaned wth the dervatve-free algorthm as we do not have an analytc expresson. 1 # c ) e s t m a c o m t j a n a n t b o o t s t r a p parametrc 2 # assumm l e s p r o b a b l t a t s son l e s que hem t r o b a t ( nomes tenm l a mostra ) 3 B0 = X$L BFGS B [ 3 ] 4 B1 = X$L BFGS B [ 4 ] 5 6 s e t. s e e d (7234) 7 s. B = p1 < exp(b0+b1 7)/(1+exp(B0+B1 7 ) ) 10 p2 < exp(b0+b1 14)/(1+exp(B0+B1 1 4 ) ) 11 p3 < exp(b0+b1 27)/(1+exp(B0+B1 2 7 ) ) 12 p4 < exp(b0+b1 57)/(1+exp(B0+B1 5 7 ) ) 6

7 13 14 r e s. b < r e p l c a t e ( s.b, 15 { 16 # e s t m a c o per metode b o o t s t rap parametrc 17 x1.b < sum(rbnom ( 5 5, 1, p1 ) ) 18 x2.b < sum(rbnom ( 1 5 7, 1, p2 ) ) 19 x3.b < sum(rbnom ( 1 5 9, 1, p3 ) ) 20 x4.b < sum(rbnom ( 1 6, 1, p4 ) ) opt.b < optm( par = c ( 0, 0 ), fn = l o g l l, x1=x1.b, x2=x2.b, x3=x3.b, x4=x4.b ) 23 opt.b$par 24 } 25 ) mean( r e s. b [ 1, ] ) 28 mean( r e s. b [ 2, ] ) e s t m a t e. v a r. b < matrx ( rep ( 0, 4 ),nrow=2) 31 e s t m a t e. v a r. b [ 1, 1 ] < var ( r e s. b [ 1, ] ) 32 e s t m a t e. v a r. b [ 2, 2 ] < var ( r e s. b [ 2, ] ) 33 e s t m a t e. v a r. b [ 1, 2 ] < cov ( r e s. b [ 1, ], r e s. b [ 2, ] ) 34 e s t m a t e. v a r. b [ 2, 1 ] < cov ( r e s. b [ 1, ], r e s. b [ 2, ] ) # S o l u t o n 37 > e s t m a t e. v a r. b 38 [, 1 ] [, 2 ] 39 [ 1, ] [ 2, ] The covarance matrx obtaned s smlar to the result obtaned n the prevous secton. Bootstrap s a possblty of estmatng the varance of the maxmum lkelhood estmators f the gradent s unknown or very dffcult to compute. It has to be taken nto account that the number of functon evaluatons of the dervatve-free method s multpled by the number of bootstrap samples. 3 Posson GLM assumpton 3.1 Algorthm deducton The densty functon of a Posson dstrbuton s Then, the lkelyhood functon can be derved as And the log-lkelhood functon L = f x l = = λx x! (6) e λ n ( λ x ) log x! = e λ λ x x! e λ (7) x log(λ ) log(x!) λ (8) If we substtute L = x (B 0 + B 1 n + B 2 z ) log(x!) e B0+B1 n+b2 z (9) 7

8 We then derve n order to get the gradent L B 1 = L B 2 = L B 0 = x x n x z e B0+B1 n+b2 z = e B0+B1 n+b2 z n = e B0+B1 n+b2 z z = x λ (10) x n x z λ n (11) λ z (12) Ths gradent can be expressed n a matrx form as x λ 1 L = n 1 n 2 n 3 n 4 x 2 λ 2 = X [x λ] (13) x z 1 z 2 z 3 z 4 3 λ 3 x 4 λ 4 And the Fsher s Informaton matrx λ λ I x = var( L) = X X t = X W X t (14) 0 0 λ λ 4 Now we formulate an teraton of the Fsher Scorng method B0 k+1 B1 k+1 B k+1 2 = And after some algebra and groupng terms we obtan When we have a model of the form B k 0 B k 1 B k 2 + (X W Xt ) 1 X W (W 1 (x λ)) (15) (X W X t ) 1 (X W ) ((X t B) + (W 1 (x λ))) (16) Y = X β + ε (17) where the varance of the error s not constant t s called Weghted Least Squares. The maxmum lkelyhood estmator of β n those models s β ML = (X t W X t ) 1 X t W Y (18) Hence, f we make a varable change we can solve an teraton of the Fsher Scorng algorthm solvng a WLS Z = ((X t B) + (W 1 (x λ))) (19) z = B 0 + n B 1 + z B 2 + x λ λ (20) 8

9 3.2 Algorthm mplementaton In ths secton the algorthm s mplemented n R followng the defntons of the prevous secton. 1 # Algorsme 2 B 0 < NULL 3 B 1 < NULL 4 B 2 < NULL 5 6 B 0 [ 1 ] = 1 7 B 1 [ 1 ] = B 2 [ 1 ] = n. t e r = lambda < matrx ( 0,nrow=n. t e r, ncol=4) 13 z < matrx ( 0,nrow=n. t e r, ncol=4) 14 v < matrx ( 0,nrow=n. t e r, ncol=4) for ( k n 1 : n. t e r ){ lambda [ k, ] = exp(b 0 [ k]+n B 1 [ k]+ z t B 2 [ k ] ) 19 z [ k, ] = B 0 [ k ] + B 1 [ k ] n + B 2 [ k ] z t + ( x lambda [ k, ] ) /lambda [ k, ] 20 v [ k, ] = lambda [ k, ] z p < z [ k, ] 23 v p < v [ k, ] 24 model < lm( z p n+z t, weghts = v p ) B 0 [ k+1] = model$ c o e f f c e n t s [ 1 ] 27 B 1 [ k+1] = model$ c o e f f c e n t s [ 2 ] 28 B 2 [ k+1] = model$ c o e f f c e n t s [ 3 ] 29 } 3.3 Algorthm executon 1 # S o l u t o n o f the algorthm 2 > lambda 3 [, 1 ] [, 2 ] [, 3 ] [, 4 ] 4 [ 1, ] [ 2, ] [ 3, ] [ 4, ] [ 5, ] [ 6, ] [ 7, ] [ 8, ] [ 9, ] [ 1 0, ] > cbnd (B 0,B 1,B 2) 16 B 0 B 1 B 2 17 [ 1, ] [ 2, ] [ 3, ] [ 4, ] [ 5, ] [ 6, ] [ 7, ]

10 24 [ 8, ] [ 9, ] [ 1 0, ] # r e s u l t a t s de l a f u n c o glm 29 glm( x n + z t, famly = posson ) # S o l u t o n > glm( x n + z t, famly=posson ) C o e f f c e n t s : 36 ( I n t e r c e p t ) n z t After 8 teratons an optmum s attaned. The result s the same wth our mplementaton and wth the functon glm of R. 10

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification Lecture 4: More classfers and classes C4B Machne Learnng Hlary 20 A. Zsserman Logstc regresson Loss functons revsted Adaboost Loss functons revsted Optmzaton Multple class classfcaton Logstc Regresson

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements Lecture 3 Densty estmaton Mlos Hauskrecht mlos@cs.ptt.edu 5329 Sennott Square Next lecture: Matlab tutoral Announcements Rules for attendng the class: Regstered for credt Regstered for audt (only f there

More information

Logistic Regression. Steve Kroon

Logistic Regression. Steve Kroon Logstc Regresson Steve Kroon Course notes sectons: 24.3-24.4 Dsclamer: these notes do not explctly ndcate whether values are vectors or scalars, but expects the reader to dscern ths from the context. Scenaro

More information

L10: Linear discriminants analysis

L10: Linear discriminants analysis L0: Lnear dscrmnants analyss Lnear dscrmnant analyss, two classes Lnear dscrmnant analyss, C classes LDA vs. PCA Lmtatons of LDA Varants of LDA Other dmensonalty reducton methods CSCE 666 Pattern Analyss

More information

Binomial Link Functions. Lori Murray, Phil Munz

Binomial Link Functions. Lori Murray, Phil Munz Bnomal Lnk Functons Lor Murray, Phl Munz Bnomal Lnk Functons Logt Lnk functon: ( p) p ln 1 p Probt Lnk functon: ( p) 1 ( p) Complentary Log Log functon: ( p) ln( ln(1 p)) Motvatng Example A researcher

More information

How To Calculate The Accountng Perod Of Nequalty

How To Calculate The Accountng Perod Of Nequalty Inequalty and The Accountng Perod Quentn Wodon and Shlomo Ytzha World Ban and Hebrew Unversty September Abstract Income nequalty typcally declnes wth the length of tme taen nto account for measurement.

More information

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by 6 CHAPTER 8 COMPLEX VECTOR SPACES 5. Fnd the kernel of the lnear transformaton gven n Exercse 5. In Exercses 55 and 56, fnd the mage of v, for the ndcated composton, where and are gven by the followng

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

Latent Class Regression. Statistics for Psychosocial Research II: Structural Models December 4 and 6, 2006

Latent Class Regression. Statistics for Psychosocial Research II: Structural Models December 4 and 6, 2006 Latent Class Regresson Statstcs for Psychosocal Research II: Structural Models December 4 and 6, 2006 Latent Class Regresson (LCR) What s t and when do we use t? Recall the standard latent class model

More information

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6

NPAR TESTS. One-Sample Chi-Square Test. Cell Specification. Observed Frequencies 1O i 6. Expected Frequencies 1EXP i 6 PAR TESTS If a WEIGHT varable s specfed, t s used to replcate a case as many tmes as ndcated by the weght value rounded to the nearest nteger. If the workspace requrements are exceeded and samplng has

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

Fisher Markets and Convex Programs

Fisher Markets and Convex Programs Fsher Markets and Convex Programs Nkhl R. Devanur 1 Introducton Convex programmng dualty s usually stated n ts most general form, wth convex objectve functons and convex constrants. (The book by Boyd and

More information

8 Algorithm for Binary Searching in Trees

8 Algorithm for Binary Searching in Trees 8 Algorthm for Bnary Searchng n Trees In ths secton we present our algorthm for bnary searchng n trees. A crucal observaton employed by the algorthm s that ths problem can be effcently solved when the

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

v a 1 b 1 i, a 2 b 2 i,..., a n b n i.

v a 1 b 1 i, a 2 b 2 i,..., a n b n i. SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 455 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces we have studed thus far n the text are real vector spaces snce the scalars are

More information

Application of Quasi Monte Carlo methods and Global Sensitivity Analysis in finance

Application of Quasi Monte Carlo methods and Global Sensitivity Analysis in finance Applcaton of Quas Monte Carlo methods and Global Senstvty Analyss n fnance Serge Kucherenko, Nlay Shah Imperal College London, UK skucherenko@mperalacuk Daro Czraky Barclays Captal DaroCzraky@barclayscaptalcom

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

BERNSTEIN POLYNOMIALS

BERNSTEIN POLYNOMIALS On-Lne Geometrc Modelng Notes BERNSTEIN POLYNOMIALS Kenneth I. Joy Vsualzaton and Graphcs Research Group Department of Computer Scence Unversty of Calforna, Davs Overvew Polynomals are ncredbly useful

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching)

Face Verification Problem. Face Recognition Problem. Application: Access Control. Biometric Authentication. Face Verification (1:1 matching) Face Recognton Problem Face Verfcaton Problem Face Verfcaton (1:1 matchng) Querymage face query Face Recognton (1:N matchng) database Applcaton: Access Control www.vsage.com www.vsoncs.com Bometrc Authentcaton

More information

Single and multiple stage classifiers implementing logistic discrimination

Single and multiple stage classifiers implementing logistic discrimination Sngle and multple stage classfers mplementng logstc dscrmnaton Hélo Radke Bttencourt 1 Dens Alter de Olvera Moraes 2 Vctor Haertel 2 1 Pontfíca Unversdade Católca do Ro Grande do Sul - PUCRS Av. Ipranga,

More information

Downlink Power Allocation for Multi-class. Wireless Systems

Downlink Power Allocation for Multi-class. Wireless Systems Downlnk Power Allocaton for Mult-class 1 Wreless Systems Jang-Won Lee, Rav R. Mazumdar, and Ness B. Shroff School of Electrcal and Computer Engneerng Purdue Unversty West Lafayette, IN 47907, USA {lee46,

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008

Risk-based Fatigue Estimate of Deep Water Risers -- Course Project for EM388F: Fracture Mechanics, Spring 2008 Rsk-based Fatgue Estmate of Deep Water Rsers -- Course Project for EM388F: Fracture Mechancs, Sprng 2008 Chen Sh Department of Cvl, Archtectural, and Envronmental Engneerng The Unversty of Texas at Austn

More information

Loop Parallelization

Loop Parallelization - - Loop Parallelzaton C-52 Complaton steps: nested loops operatng on arrays, sequentell executon of teraton space DECLARE B[..,..+] FOR I :=.. FOR J :=.. I B[I,J] := B[I-,J]+B[I-,J-] ED FOR ED FOR analyze

More information

Characterization of Assembly. Variation Analysis Methods. A Thesis. Presented to the. Department of Mechanical Engineering. Brigham Young University

Characterization of Assembly. Variation Analysis Methods. A Thesis. Presented to the. Department of Mechanical Engineering. Brigham Young University Characterzaton of Assembly Varaton Analyss Methods A Thess Presented to the Department of Mechancal Engneerng Brgham Young Unversty In Partal Fulfllment of the Requrements for the Degree Master of Scence

More information

Pricing Multi-Asset Cross Currency Options

Pricing Multi-Asset Cross Currency Options CIRJE-F-844 Prcng Mult-Asset Cross Currency Optons Kenchro Shraya Graduate School of Economcs, Unversty of Tokyo Akhko Takahash Unversty of Tokyo March 212; Revsed n September, October and November 212

More information

Regression Models for a Binary Response Using EXCEL and JMP

Regression Models for a Binary Response Using EXCEL and JMP SEMATECH 997 Statstcal Methods Symposum Austn Regresson Models for a Bnary Response Usng EXCEL and JMP Davd C. Trndade, Ph.D. STAT-TECH Consultng and Tranng n Appled Statstcs San Jose, CA Topcs Practcal

More information

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems Jont Schedulng of Processng and Shuffle Phases n MapReduce Systems Fangfe Chen, Mural Kodalam, T. V. Lakshman Department of Computer Scence and Engneerng, The Penn State Unversty Bell Laboratores, Alcatel-Lucent

More information

On the Solution of Indefinite Systems Arising in Nonlinear Optimization

On the Solution of Indefinite Systems Arising in Nonlinear Optimization On the Soluton of Indefnte Systems Arsng n Nonlnear Optmzaton Slva Bonettn, Valera Ruggero and Federca Tnt Dpartmento d Matematca, Unverstà d Ferrara Abstract We consder the applcaton of the precondtoned

More information

GENETIC ALGORITHM FOR PROJECT SCHEDULING AND RESOURCE ALLOCATION UNDER UNCERTAINTY

GENETIC ALGORITHM FOR PROJECT SCHEDULING AND RESOURCE ALLOCATION UNDER UNCERTAINTY Int. J. Mech. Eng. & Rob. Res. 03 Fady Safwat et al., 03 Research Paper ISS 78 049 www.jmerr.com Vol., o. 3, July 03 03 IJMERR. All Rghts Reserved GEETIC ALGORITHM FOR PROJECT SCHEDULIG AD RESOURCE ALLOCATIO

More information

Implementation of Deutsch's Algorithm Using Mathcad

Implementation of Deutsch's Algorithm Using Mathcad Implementaton of Deutsch's Algorthm Usng Mathcad Frank Roux The followng s a Mathcad mplementaton of Davd Deutsch's quantum computer prototype as presented on pages - n "Machnes, Logc and Quantum Physcs"

More information

A Lyapunov Optimization Approach to Repeated Stochastic Games

A Lyapunov Optimization Approach to Repeated Stochastic Games PROC. ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING, OCT. 2013 1 A Lyapunov Optmzaton Approach to Repeated Stochastc Games Mchael J. Neely Unversty of Southern Calforna http://www-bcf.usc.edu/

More information

THE METHOD OF LEAST SQUARES THE METHOD OF LEAST SQUARES

THE METHOD OF LEAST SQUARES THE METHOD OF LEAST SQUARES The goal: to measure (determne) an unknown quantty x (the value of a RV X) Realsaton: n results: y 1, y 2,..., y j,..., y n, (the measured values of Y 1, Y 2,..., Y j,..., Y n ) every result s encumbered

More information

Lecture 3: Force of Interest, Real Interest Rate, Annuity

Lecture 3: Force of Interest, Real Interest Rate, Annuity Lecture 3: Force of Interest, Real Interest Rate, Annuty Goals: Study contnuous compoundng and force of nterest Dscuss real nterest rate Learn annuty-mmedate, and ts present value Study annuty-due, and

More information

MARKET SHARE CONSTRAINTS AND THE LOSS FUNCTION IN CHOICE BASED CONJOINT ANALYSIS

MARKET SHARE CONSTRAINTS AND THE LOSS FUNCTION IN CHOICE BASED CONJOINT ANALYSIS MARKET SHARE CONSTRAINTS AND THE LOSS FUNCTION IN CHOICE BASED CONJOINT ANALYSIS Tmothy J. Glbrde Assstant Professor of Marketng 315 Mendoza College of Busness Unversty of Notre Dame Notre Dame, IN 46556

More information

1 Example 1: Axis-aligned rectangles

1 Example 1: Axis-aligned rectangles COS 511: Theoretcal Machne Learnng Lecturer: Rob Schapre Lecture # 6 Scrbe: Aaron Schld February 21, 2013 Last class, we dscussed an analogue for Occam s Razor for nfnte hypothess spaces that, n conjuncton

More information

An interactive system for structure-based ASCII art creation

An interactive system for structure-based ASCII art creation An nteractve system for structure-based ASCII art creaton Katsunor Myake Henry Johan Tomoyuk Nshta The Unversty of Tokyo Nanyang Technologcal Unversty Abstract Non-Photorealstc Renderng (NPR), whose am

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background:

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background: SPEE Recommended Evaluaton Practce #6 efnton of eclne Curve Parameters Background: The producton hstores of ol and gas wells can be analyzed to estmate reserves and future ol and gas producton rates and

More information

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP)

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP) 6.3 / -- Communcaton Networks II (Görg) SS20 -- www.comnets.un-bremen.de Communcaton Networks II Contents. Fundamentals of probablty theory 2. Emergence of communcaton traffc 3. Stochastc & Markovan Processes

More information

On fourth order simultaneously zero-finding method for multiple roots of complex polynomial equations 1

On fourth order simultaneously zero-finding method for multiple roots of complex polynomial equations 1 General Mathematcs Vol. 6, No. 3 (2008), 9 3 On fourth order smultaneously zero-fndng method for multple roots of complex polynomal euatons Nazr Ahmad Mr and Khald Ayub Abstract In ths paper, we present

More information

An MILP model for planning of batch plants operating in a campaign-mode

An MILP model for planning of batch plants operating in a campaign-mode An MILP model for plannng of batch plants operatng n a campagn-mode Yanna Fumero Insttuto de Desarrollo y Dseño CONICET UTN yfumero@santafe-concet.gov.ar Gabrela Corsano Insttuto de Desarrollo y Dseño

More information

General Iteration Algorithm for Classification Ratemaking

General Iteration Algorithm for Classification Ratemaking General Iteraton Algorthm for Classfcaton Ratemakng by Luyang Fu and Cheng-sheng eter Wu ABSTRACT In ths study, we propose a flexble and comprehensve teraton algorthm called general teraton algorthm (GIA)

More information

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network 700 Proceedngs of the 8th Internatonal Conference on Innovaton & Management Forecastng the Demand of Emergency Supples: Based on the CBR Theory and BP Neural Network Fu Deqang, Lu Yun, L Changbng School

More information

Lecture 5,6 Linear Methods for Classification. Summary

Lecture 5,6 Linear Methods for Classification. Summary Lecture 5,6 Lnear Methods for Classfcaton Rce ELEC 697 Farnaz Koushanfar Fall 2006 Summary Bayes Classfers Lnear Classfers Lnear regresson of an ndcator matrx Lnear dscrmnant analyss (LDA) Logstc regresson

More information

Inverse Modeling of Tight Gas Reservoirs

Inverse Modeling of Tight Gas Reservoirs Inverse Modelng of Tght Gas Reservors Der Fakultät für Geowssenschaften, Geotechnk und Bergbau der Technschen Unverstät Bergakademe Freberg engerechte Dssertaton Zur Erlangung des akademschen Grades Doktor-Ingeneur

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

APPLICATION OF COMPUTER PROGRAMMING IN OPTIMIZATION OF TECHNOLOGICAL OBJECTIVES OF COLD ROLLING

APPLICATION OF COMPUTER PROGRAMMING IN OPTIMIZATION OF TECHNOLOGICAL OBJECTIVES OF COLD ROLLING Journal Journal of Chemcal of Chemcal Technology and and Metallurgy, 50, 6, 50, 2015, 6, 2015 638-643 APPLICATION OF COMPUTER PROGRAMMING IN OPTIMIZATION OF TECHNOLOGICAL OBJECTIVES OF COLD ROLLING Abdrakhman

More information

A fast method for binary programming using first-order derivatives, with application to topology optimization with buckling constraints

A fast method for binary programming using first-order derivatives, with application to topology optimization with buckling constraints INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING Int. J. Numer. Meth. Engng (2012) Publshed onlne n Wley Onlne Lbrary (wleyonlnelbrary.com)..4367 A fast method for bnary programmng usng frst-order

More information

A Practitioner's Guide to Generalized Linear Models

A Practitioner's Guide to Generalized Linear Models A Practtoner's Gude to Generalzed Lnear Models A CAS Study Note Duncan Anderson, FIA Sholom Feldblum, FCAS Claudne Modln, FCAS Dors Schrmacher, FCAS Ernesto Schrmacher, ASA Neeza Thand, FCAS Thrd Edton

More information

Imperial College London

Imperial College London F. Fang 1, C.C. Pan 1, I.M. Navon 2, M.D. Pggott 1, G.J. Gorman 1, P.A. Allson 1 and A.J.H. Goddard 1 1 Appled Modellng and Computaton Group Department of Earth Scence and Engneerng Imperal College London,

More information

The OC Curve of Attribute Acceptance Plans

The OC Curve of Attribute Acceptance Plans The OC Curve of Attrbute Acceptance Plans The Operatng Characterstc (OC) curve descrbes the probablty of acceptng a lot as a functon of the lot s qualty. Fgure 1 shows a typcal OC Curve. 10 8 6 4 1 3 4

More information

An efficient constraint handling methodology for multi-objective evolutionary algorithms

An efficient constraint handling methodology for multi-objective evolutionary algorithms Rev. Fac. Ing. Unv. Antoqua N. 49. pp. 141-150. Septembre, 009 An effcent constrant handlng methodology for mult-objectve evolutonary algorthms Una metodología efcente para manejo de restrccones en algortmos

More information

Production. 2. Y is closed A set is closed if it contains its boundary. We need this for the solution existence in the profit maximization problem.

Production. 2. Y is closed A set is closed if it contains its boundary. We need this for the solution existence in the profit maximization problem. Producer Theory Producton ASSUMPTION 2.1 Propertes of the Producton Set The producton set Y satsfes the followng propertes 1. Y s non-empty If Y s empty, we have nothng to talk about 2. Y s closed A set

More information

Risk Model of Long-Term Production Scheduling in Open Pit Gold Mining

Risk Model of Long-Term Production Scheduling in Open Pit Gold Mining Rsk Model of Long-Term Producton Schedulng n Open Pt Gold Mnng R Halatchev 1 and P Lever 2 ABSTRACT Open pt gold mnng s an mportant sector of the Australan mnng ndustry. It uses large amounts of nvestments,

More information

This paper can be downloaded without charge from the Social Sciences Research Network Electronic Paper Collection: http://ssrn.com/abstract=2694633

This paper can be downloaded without charge from the Social Sciences Research Network Electronic Paper Collection: http://ssrn.com/abstract=2694633 Workng Paper Coordnatng Prcng and Inventory Replenshment wth Nonparametrc Demand Learnng Boxao Chen Department of Industral and Operatons Engneerng Unversty of Mchgan Xul Chao Department of Industral and

More information

Chapter 2 The Basics of Pricing with GLMs

Chapter 2 The Basics of Pricing with GLMs Chapter 2 The Bascs of Prcng wth GLMs As descrbed n the prevous secton, the goal of a tarff analyss s to determne how one or more key ratos Y vary wth a number of ratng factors Ths s remnscent of analyzng

More information

GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM

GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM BARRIOT Jean-Perre, SARRAILH Mchel BGI/CNES 18.av.E.Beln 31401 TOULOUSE Cedex 4 (France) Emal: jean-perre.barrot@cnes.fr 1/Introducton The

More information

Credit Limit Optimization (CLO) for Credit Cards

Credit Limit Optimization (CLO) for Credit Cards Credt Lmt Optmzaton (CLO) for Credt Cards Vay S. Desa CSCC IX, Ednburgh September 8, 2005 Copyrght 2003, SAS Insttute Inc. All rghts reserved. SAS Propretary Agenda Background Tradtonal approaches to credt

More information

An Enhanced Super-Resolution System with Improved Image Registration, Automatic Image Selection, and Image Enhancement

An Enhanced Super-Resolution System with Improved Image Registration, Automatic Image Selection, and Image Enhancement An Enhanced Super-Resoluton System wth Improved Image Regstraton, Automatc Image Selecton, and Image Enhancement Yu-Chuan Kuo ( ), Chen-Yu Chen ( ), and Chou-Shann Fuh ( ) Department of Computer Scence

More information

1. Measuring association using correlation and regression

1. Measuring association using correlation and regression How to measure assocaton I: Correlaton. 1. Measurng assocaton usng correlaton and regresson We often would lke to know how one varable, such as a mother's weght, s related to another varable, such as a

More information

Statistical Methods to Develop Rating Models

Statistical Methods to Develop Rating Models Statstcal Methods to Develop Ratng Models [Evelyn Hayden and Danel Porath, Österrechsche Natonalbank and Unversty of Appled Scences at Manz] Source: The Basel II Rsk Parameters Estmaton, Valdaton, and

More information

Quantization Effects in Digital Filters

Quantization Effects in Digital Filters Quantzaton Effects n Dgtal Flters Dstrbuton of Truncaton Errors In two's complement representaton an exact number would have nfntely many bts (n general). When we lmt the number of bts to some fnte value

More information

PRACTICE 1: MUTUAL FUNDS EVALUATION USING MATLAB.

PRACTICE 1: MUTUAL FUNDS EVALUATION USING MATLAB. PRACTICE 1: MUTUAL FUNDS EVALUATION USING MATLAB. INDEX 1. Load data usng the Edtor wndow and m-fle 2. Learnng to save results from the Edtor wndow. 3. Computng the Sharpe Rato 4. Obtanng the Treynor Rato

More information

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12 14 The Ch-squared dstrbuton PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 1 If a normal varable X, havng mean µ and varance σ, s standardsed, the new varable Z has a mean 0 and varance 1. When ths standardsed

More information

Optimal monitoring in large networks by Successive c-optimal Designs

Optimal monitoring in large networks by Successive c-optimal Designs Optmal montorng n large networs by Successve c-optmal Desgns Gullaume Sagnol, Stéphane Gaubert INRIA Saclay & CMAP, Ecole Polytechnque Emal: {gullaumesagnol,stephanegaubert}@nrafr Mustapha Bouhtou Orange

More information

Harder, Better, Faster, Stronger Convergence Rates for Least-Squares Regression

Harder, Better, Faster, Stronger Convergence Rates for Least-Squares Regression Harder, Better, Faster, Stronger Convergence Rates for Least-Squares Regresson Aymerc Deuleveut, Ncolas Flammaron, Francs Bach To cte ths verson: Aymerc Deuleveut, Ncolas Flammaron, Francs Bach. Harder,

More information

Can Auto Liability Insurance Purchases Signal Risk Attitude?

Can Auto Liability Insurance Purchases Signal Risk Attitude? Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159-164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? Chu-Shu L Department of Internatonal Busness, Asa Unversty, Tawan Sheng-Chang

More information

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and POLYSA: A Polynomal Algorthm for Non-bnary Constrant Satsfacton Problems wth and Mguel A. Saldo, Federco Barber Dpto. Sstemas Informátcos y Computacón Unversdad Poltécnca de Valenca, Camno de Vera s/n

More information

Georey E. Hinton. University oftoronto. Email: zoubin@cs.toronto.edu. Technical Report CRG-TR-96-1. May 21, 1996 (revised Feb 27, 1997) Abstract

Georey E. Hinton. University oftoronto. Email: zoubin@cs.toronto.edu. Technical Report CRG-TR-96-1. May 21, 1996 (revised Feb 27, 1997) Abstract The EM Algorthm for Mxtures of Factor Analyzers Zoubn Ghahraman Georey E. Hnton Department of Computer Scence Unversty oftoronto 6 Kng's College Road Toronto, Canada M5S A4 Emal: zoubn@cs.toronto.edu Techncal

More information

Research Article Enhanced Two-Step Method via Relaxed Order of α-satisfactory Degrees for Fuzzy Multiobjective Optimization

Research Article Enhanced Two-Step Method via Relaxed Order of α-satisfactory Degrees for Fuzzy Multiobjective Optimization Hndaw Publshng Corporaton Mathematcal Problems n Engneerng Artcle ID 867836 pages http://dxdoorg/055/204/867836 Research Artcle Enhanced Two-Step Method va Relaxed Order of α-satsfactory Degrees for Fuzzy

More information

Approximating Cross-validatory Predictive Evaluation in Bayesian Latent Variables Models with Integrated IS and WAIC

Approximating Cross-validatory Predictive Evaluation in Bayesian Latent Variables Models with Integrated IS and WAIC Approxmatng Cross-valdatory Predctve Evaluaton n Bayesan Latent Varables Models wth Integrated IS and WAIC Longha L Department of Mathematcs and Statstcs Unversty of Saskatchewan Saskatoon, SK, CANADA

More information

On the Interaction between Load Balancing and Speed Scaling

On the Interaction between Load Balancing and Speed Scaling On the Interacton between Load Balancng and Speed Scalng Ljun Chen, Na L and Steven H. Low Engneerng & Appled Scence Dvson, Calforna Insttute of Technology, USA Abstract Speed scalng has been wdely adopted

More information

Testing and Debugging Resource Allocation for Fault Detection and Removal Process

Testing and Debugging Resource Allocation for Fault Detection and Removal Process Internatonal Journal of New Computer Archtectures and ther Applcatons (IJNCAA) 4(4): 93-00 The Socety of Dgtal Informaton and Wreless Communcatons, 04 (ISSN: 0-9085) Testng and Debuggng Resource Allocaton

More information

Alternate Approximation of Concave Cost Functions for

Alternate Approximation of Concave Cost Functions for Alternate Approxmaton of Concave Cost Functons for Process Desgn and Supply Chan Optmzaton Problems Dego C. Cafaro * and Ignaco E. Grossmann INTEC (UNL CONICET), Güemes 3450, 3000 Santa Fe, ARGENTINA Department

More information

Everybody needs somebody: Modeling social and grouping behavior on a linear programming multiple people tracker

Everybody needs somebody: Modeling social and grouping behavior on a linear programming multiple people tracker Everybody needs somebody: Modelng socal and groupng behavor on a lnear programmng multple people tracker Laura Leal-Taxe, Gerard Pons-Moll and Bodo Rosenhahn Insttute for Informaton Processng (TNT) Lebnz

More information

SUMMARY. Topology optimization, buckling, eigenvalue, derivative, structural optimization 1. INTRODUCTION

SUMMARY. Topology optimization, buckling, eigenvalue, derivative, structural optimization 1. INTRODUCTION INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING Int. J. Numer. Meth. Engng 0000; 00:1 20 Publshed onlne n Wley InterScence (www.nterscence.wley.com). A fast method for solvng bnary programmng

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

Chapter XX More advanced approaches to the analysis of survey data. Gad Nathan Hebrew University Jerusalem, Israel. Abstract

Chapter XX More advanced approaches to the analysis of survey data. Gad Nathan Hebrew University Jerusalem, Israel. Abstract Household Sample Surveys n Developng and Transton Countres Chapter More advanced approaches to the analyss of survey data Gad Nathan Hebrew Unversty Jerusalem, Israel Abstract In the present chapter, we

More information

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks Energy Effcent Routng n Ad Hoc Dsaster Recovery Networks Gl Zussman and Adran Segall Department of Electrcal Engneerng Technon Israel Insttute of Technology Hafa 32000, Israel {glz@tx, segall@ee}.technon.ac.l

More information

Exhaustive Regression. An Exploration of Regression-Based Data Mining Techniques Using Super Computation

Exhaustive Regression. An Exploration of Regression-Based Data Mining Techniques Using Super Computation Exhaustve Regresson An Exploraton of Regresson-Based Data Mnng Technques Usng Super Computaton Antony Daves, Ph.D. Assocate Professor of Economcs Duquesne Unversty Pttsburgh, PA 58 Research Fellow The

More information

Real-time adaptive information-theoretic optimization of neurophysiology experiments

Real-time adaptive information-theoretic optimization of neurophysiology experiments Real-tme adaptve nformaton-theoretc optmzaton of neurophysology experments Jeremy Lew School of Boengneerng Georga Insttute of Technology jlew@gatech.edu Robert Butera School of Electrcal and Computer

More information

Fragility Based Rehabilitation Decision Analysis

Fragility Based Rehabilitation Decision Analysis .171. Fraglty Based Rehabltaton Decson Analyss Cagdas Kafal Graduate Student, School of Cvl and Envronmental Engneerng, Cornell Unversty Research Supervsor: rcea Grgoru, Professor Summary A method s presented

More information

Fast gradient descent method for mean-cvar optimization

Fast gradient descent method for mean-cvar optimization Fast gradent descent method for mean-cvar optmzaton Garud Iyengar Alfred Ka Chun Ma February 27, 2009 Abstract We propose an teratve gradent descent procedure for computng approxmate solutons for the scenaro-based

More information

Automated information technology for ionosphere monitoring of low-orbit navigation satellite signals

Automated information technology for ionosphere monitoring of low-orbit navigation satellite signals Automated nformaton technology for onosphere montorng of low-orbt navgaton satellte sgnals Alexander Romanov, Sergey Trusov and Alexey Romanov Federal State Untary Enterprse Russan Insttute of Space Devce

More information

STATISTICAL DATA ANALYSIS IN EXCEL

STATISTICAL DATA ANALYSIS IN EXCEL Mcroarray Center STATISTICAL DATA ANALYSIS IN EXCEL Lecture 6 Some Advanced Topcs Dr. Petr Nazarov 14-01-013 petr.nazarov@crp-sante.lu Statstcal data analyss n Ecel. 6. Some advanced topcs Correcton for

More information

Least Squares Fitting of Data

Least Squares Fitting of Data Least Squares Fttng of Data Davd Eberly Geoetrc Tools, LLC http://www.geoetrctools.co/ Copyrght c 1998-2016. All Rghts Reserved. Created: July 15, 1999 Last Modfed: January 5, 2015 Contents 1 Lnear Fttng

More information

ONLINE ANOMALY DETECTION WITH CONSTANT FALSE ALARM RATE

ONLINE ANOMALY DETECTION WITH CONSTANT FALSE ALARM RATE 2015 IEEE INTERNATIONAL WORKSHOP ON MACHINE LEARNING FOR SIGNAL PROCESSING, SEPT. 17 20, 2014, BOSTON, USA ONLINE ANOMALY DETECTION WITH CONSTANT FALSE ALARM RATE Huseyn Ozkan Fath Ozkan Ibrahm Delbalta

More information

1.2 DISTRIBUTIONS FOR CATEGORICAL DATA

1.2 DISTRIBUTIONS FOR CATEGORICAL DATA DISTRIBUTIONS FOR CATEGORICAL DATA 5 present models for a categorcal response wth matched pars; these apply, for nstance, wth a categorcal response measured for the same subjects at two tmes. Chapter 11

More information

SIMULATION OF INVENTORY CONTROL SYSTEM FOR SUPPLY CHAIN PRODUCER WHOLESALER CLIENT IN EXTENDSIM ENVIRONMENT

SIMULATION OF INVENTORY CONTROL SYSTEM FOR SUPPLY CHAIN PRODUCER WHOLESALER CLIENT IN EXTENDSIM ENVIRONMENT SIMULATION OF INVENTOY CONTOL SYSTEM FO SUPPLY CHAIN PODUCE WHOLESALE CLIENT IN EXTENDSIM ENVIONMENT Eugene Kopytov and Avars Muravjovs Transport and Telecommuncaton Insttute, Lomonosov Street, ga, LV-09,

More information

A machine vision approach for detecting and inspecting circular parts

A machine vision approach for detecting and inspecting circular parts A machne vson approach for detectng and nspectng crcular parts Du-Mng Tsa Machne Vson Lab. Department of Industral Engneerng and Management Yuan-Ze Unversty, Chung-L, Tawan, R.O.C. E-mal: edmtsa@saturn.yzu.edu.tw

More information

1 De nitions and Censoring

1 De nitions and Censoring De ntons and Censorng. Survval Analyss We begn by consderng smple analyses but we wll lead up to and take a look at regresson on explanatory factors., as n lnear regresson part A. The mportant d erence

More information

Realistic Image Synthesis

Realistic Image Synthesis Realstc Image Synthess - Combned Samplng and Path Tracng - Phlpp Slusallek Karol Myszkowsk Vncent Pegoraro Overvew: Today Combned Samplng (Multple Importance Samplng) Renderng and Measurng Equaton Random

More information

Online Advertisement, Optimization and Stochastic Networks

Online Advertisement, Optimization and Stochastic Networks Onlne Advertsement, Optmzaton and Stochastc Networks Bo (Rambo) Tan and R. Srkant Department of Electrcal and Computer Engneerng Unversty of Illnos at Urbana-Champagn Urbana, IL, USA 1 arxv:1009.0870v6

More information

Method for Production Planning and Inventory Control in Oil

Method for Production Planning and Inventory Control in Oil Memors of the Faculty of Engneerng, Okayama Unversty, Vol.41, pp.20-30, January, 2007 Method for Producton Plannng and Inventory Control n Ol Refnery TakujImamura,MasamKonshandJunIma Dvson of Electronc

More information