New Routes from Minimal Approximation Error to Principal Components

Size: px
Start display at page:

Download "New Routes from Minimal Approximation Error to Principal Components"

Transcription

1 Neural Process Lett (2008) 27: DOI 0.007/s New Routes from Miimal Approximatio Error to Pricipal Compoets Abhilash Alexader Mirada Ya-Aël Le Borge Gialuca Botempi Published olie: 5 Jauary 2008 Spriger Sciece+Busiess Media, LLC Abstract We itroduce two ew methods of derivig the classical PCA i the framework of miimizig the mea square error upo performig a lower-dimesioal approximatio of the data. These methods are based o two forms of the mea square error fuctio. Oe of the ovelties of the preseted methods is that the commoly employed process of subtractio of the mea of the data becomes part of the solutio of the optimizatio problem ad ot a pre-aalysis heuristic. We also derive the optimal basis ad the miimum error of approximatio i this framework ad demostrate the elegace of our solutio i compariso with a recet solutio i the framework. Keywords Pricipal compoets aalysis Eigevalue Matrix trace Itroductio The problem of approximatig a give set of data usig a weighted liear combiatio of a fewer umber of vectors tha the origial dimesioality is classic. May applicatios that require such a dimesioality reductio desire that the ew represetatio retai the maximum variability i the data for further aalysis. A popular method that attais simultaeous dimesioality reductio, miimum mea square error of approximatio ad retaimet of maximum variace of the origial data represetatio i the ew represetatio is called the Pricipal Compoets Aalysis (PCA) [7,]. The most popular framework for derivig PCA starts with the aalysis of variace. A very commo derivatio of PCA i this framework geerates the basis by iteratively fidig the orthogoal directios of maximum retaied variaces [7,0,,4]. Sice variace is implied i the statemet of the problem here, the mea is subtracted from the data as a prelimiary step. The secod most predomiat framework derives PCA by miimizig the A. A. Mirada (B) Y.-A. Le Borge G. Botempi Machie Learig Group, Départemet d Iformatique, Uiversité Libre de Bruxelles, Boulevard du Triomphe CP22, 050 Brussels, Belgium abalexa@ulb.ac.be

2 98 A. A. Mirada et al. mea square error of approximatio [ 3]. Aided by the derivatio i the variace-based framework above, it has become acceptable to resort to mea subtractio of the data prior to ay aalysis i this framework too i order to keep the aalysis simple. I this letter our focus is o the latter framework withi which we demostrate two distict ad elegat aalytical methods of derivig the PCA. I each of these methods of derivatio, subtractio of data mea becomes part of the solutio istead of beig a iitial assumptio. The letter is orgaized as follows: i Sect. 2 we describe the motivatio behid the eed for yet aother derivatio of the classical PCA. I particular, we highlight the issue of mea ceterig i Sect. 2.. The otatios are itroduced i Sect. 2.2 ad the PCA problem ad its iterpretatios are discussed i Sect. 3. After reviewig a recet solutio i Sect. 4, we make it evidet i Sect. 5 that our two methods are due to two forms of the optimizatio fuctio. The we itroduce these two methods of solvig the PCA problem i Sects. 6 ad 7 ad arrive at a simple commo form of the optimizatio fuctio i both these methods. This is aalyzed further i Sect. 8 where we show the relatio of the variace to the optimal basis as well as the miimum approximatio error attaied i PCA. I Sect. 8.3, we revisit a recet solutio i our framework of PCA itroduced i Sect. 4 ad equate it with our approach. 2 Motivatio There are may stadard textbooks of multivariate ad statistical aalysis [0,,4] detailig PCA as a techique that seeks the best approximatio of a give set of data poits usig a liear combiatio of a set of vectors which retai maximum variace alog their directios. Sice this framework of PCA starts by fidig the covariaces, the mea has to be subtracted from the data ad becomes the de facto origi of the ew coordiate system. The subsequet aalysis is simple: fid the eigevector correspodig to the largest eigevalue of the covariace matrix as the first basis vector. The fid the secod basis vector o which the data compoets bear zero correlatio with the data compoets o the first basis vector. This turs out to be the eigevector correspodig to the secod largest eigevalue. I successively fidig the basis vectors that have ucorrelated compoets as the eigevectors of decreasig retaied variaces, the secod order cross momets betwee the compoets are successively elimiated. Computatioally, a widely employed trick i this framework fids the eigevectors usig sigular value decompositio of the mea cetered data matrix which effectively diagoalizes the covariace matrix without actually computig it [,4]. The set of orthogoal vectors correspodig to the largest few sigular values proportioal to the variaces yields those directios which retai the maximum variace i the ew represetatio of the data. The secod framework derives the PCA approximatio by usig its property of miimizig the mea square error. We thik that this framework is more effective i itroducig PCA to a ovice because the two outcomes of optimal dimesioality reductio, viz. error miimizatio ad retaied variace maximizatio, are attaied here simultaeously. Followig the path of the retaied variace maximizatio framework ad to keep the aalysis simple, may textbooks [2,9,0,20] advocate a mea subtractio for this framework too without sesible justificatio. Pearso stated i his ow classical paper [8]: The secod momet of a system about a series of parallel lies is always least for the lie goig through the cetroid. Hece: The best-fittig straight lie for a system of poits i a space of ay order goes through the cetroid of the system. Elimiatio of higher order cross momets is dealt i Idepedet Compoets Aalysis (ICA) [9].

3 New Routes from Miimal Approximatio Error to Pricipal Compoets 99 A procedure equivalet to rephrasig of this statemet is followed i a much refereced textbook [3] which reasos that sice the mea is the zero-dimesioal hyperplae which satisfies the miimum average square error criterio, ay higher dimesioal hyperplae should be excused to pass through it too. I order to keep our aalysis coheret with the cocept of simultaeous dimesioality reductio, retaied variace maximizatio ad approximatio error miimizatio, we do ot ivite the reader to such geometric ituitios. Note that the error miimizatio framework ca also be viewed as a total least squares regressio problem with all variables thought to be free so that the task is to fit a lower dimesioal hyperplae that miimizes the perpedicular distaces from the data poits to the hyperplae [2]. We will also be reviewig [] who derives PCA i the same framework as that of ours. Ulike i their approach, we either udertake a complete orthogoal decompositio or force ay basis vectors to bear a commo statistic eticed by the prospect of a evetual mea subtractio. Also for the beefit of practitioers who would like to deal data as realizatios of a radom variable, our treatmet i the data samples domai ca be readily exteded to a populatio domai. 2. To Mea Ceter or Not I the framework of fidig the basis of a lower dimesioal space which miimizes the mea square error of approximatio, the process of mea subtractio has so far bee part of the heuristics that the data eeds to be cetered before istallig the ew low-dimesioal coordiate system motivated by the philosophy accordig to [8] that, had the mea of the data ot bee subtracted, the best fittig hyperplae would pass through the origi ad ot through the cetroid. But there exist situatios where a hyperplae is merely expected to partitio the data space ito orthogoal subspaces ad as a result subtractio of mea is ot desired. Note that i such situatios, the term pricipal compoet does ot strictly hold as the basis vectors for the ew space are ot obtaied from the data covariace matrix ad the mai cocer there is the decompositio of the data rather tha its approximatio. Oe such set of situatios are addressed by the Fukuaga Kootz Trasform [5,6] ad it works by ot requirig a subtractio of mea but istead fids the pricipal compoets of the autocorrelatio matrices of two classes of data. It is widely used i automatic target recogitio where eigevalue decompositio geerates basis for a target space orthogoal to the clutter space. But such is the issue of mea subtractio i usig this trasform that researchers of [2] ad[8] use autocorrelatio ad covariace matrices, respectively, for the same task without a justificatio of the impact of their choice to mea ceter or ot. A similar approach called Eigespace Separatio Trasformatio [9] aimed at classificatio also does ot ivolve mea subtractio. A family of techiques called Orthogoal Subspace Projectio that is widely applied i oise rejectio of sigals use data that are ot mea cetered for the geeralized PCA that follows [6]. Although the theory of PCA demads mea subtractio for optimal low dimesioal approximatio, for may applicatios it is ot without cosequece. For example, the researchers of ecology ad climate studies have extesively debated the purpose ad result of mea ceterig for their PCA-based data aalysis. I [7], the characteristics ad apparet advatages of the pricipal compoets geerated without mea subtractio are compared for data sampled homogeously i the origial space or otherwise. The claim made therei is that if data form distict clusters, the ifluece of variace withi a cluster o aother ca be miimized by ot subtractig the mea. Aother ogoig debate amed Hockey Stick cotroversy [5] ivolves the appropriateess of mea subtractio for PCA i a much cited global warmig study [3].

4 200 A. A. Mirada et al. It should be bore i mid that this letter is either solely about the aforemetioed issue of mea ceterig that researchers usig PCA ofte take it for grated or does it chage the results of PCA that is previously kow to them. But we demostrate i a ew comprehesive framework that (i) the mea subtractio becomes a solutio to the optimizatio problem i PCA ad we reach this solutio through two simple distict methods that borrow little from traditioal textbook derivatios of PCA, ad (ii) the derivatio of the basis for the low dimesioal space coverges to miimum approximatio error ad maximum retaied variace i the framework. Cosequetly, we believe that may problems which raise questios about their choice regardig mea subtractio ca be revisited with ease usig our proposed PCA framework. 2.2 Notatios The otatios that will be used throughout this letter are summarized i the table below J q : error fuctio q : ew dimesioality p : origial dimesioality : umber of samples x k R p ; k th data sample ˆx k R p ; approximatio of x k θ R p ; ew geeral origi x k = x k θ R p e i R p ; i th orthoormal basis vector of R p W = [e e q ] R p q B = I WW T R p p W = [e q+ e p ] R p p q z k R q ; depedet o x k b R p q ; a costat Tr(A) : Trace of the matrix A rak(a) : Rak of the matrix A µ R p ; sample mea S R p p ; sample covariace matrix λ i : i th largest eigevalue of S r = rak(s) 3 Problem Defiitio i the Sample Domai Let x k R p,k =,...,be a give set of data poits. Suppose we are iterested i orthoormal vectors e i R p,i =,...,q p whose resultat of weighted liear combiatio ˆx k R p ca approximate x k with a miimum average (sample mea) square error or i other words miimize J q ( ˆx k ) = x k ˆx k 2. () The problem stated above meas that we eed a approximatio x k ˆx k such that ˆx k = q i= ( ) ei T x k e i (2) so that we attai the miimum for J q. This approximatio assumes that the origi of all orthoormal e i is the same as that of the coordiate system i which the data is defied. We assume orthoormality here because (i) orthogoality guaratees liearly idepedet e i so

5 New Routes from Miimal Approximatio Error to Pricipal Compoets 20 that they form a basis for R q (ii) ormalizig e i maitais otatioal simplicity i ot havig to divide the scalars e T i x k i (2) by the orm e i which is uity due to our assumptio. We reformulate the approximatio ˆx k = θ + q i= ( ) ei T (x k θ) e i (3) to assume that the ew represetatio usig basis vectors e i has a geeral origi θ R p ad ot the origi as i the approximatio (2). Hece, the PCA problem may be defied as ˆx k = θ + q ( argmi x k ˆx k 2 i= e T i (x k θ) ) e i ; : (4) e i,θ ei T e j = 0, i = j; ei T e i = i, j. which seeks a set of orthoormal basis vectors e i with a ew origi θ which miimizes the errorfuctioi() i order to fid a low-dimesioal approximatio W T (x k θ) R q for ay x k R p,where It is ow easy to see that (3) becomes W =[e e q ]. (5) ˆx k = θ + WW T (x k θ). (6) Hece the displacemet vector directed from the approximatio ˆx k towards x k is x k ˆx k = (x k θ) WW T (x k θ), which usig x k = x k θ ca be writte cocisely as x k ˆx k = x k WW T x k. By settig B = I WW T for simplicity of otatio, we write the displacemet vector as x k ˆx k = B x k. (7) 4 Review of a Recet Solutio The most recet PCA solutio i the framework of approximatio error miimizatio, derived i [], is reviewed here. They derive PCA by udertakig a complete decompositio ˆx k = Wz k + Wb (8) ito basis vectors cotaied i the colums of matrix W of (5) ad W =[e q+ e p ] R p p q such that compoets of z k R q deped o x k, whereas compoets of b R p q are costats commo for all data poits. By takig the derivative of the error fuctio with respect to b, they fid that b = W T µ (9) so that the commo compoets are those of the sample mea vector µ. This implies that by subtractig the sample mea they are o loger obliged to retai the p q dimesios correspodig to the colums of W which preserve little iformatio regardig the variatio i the data. The first drawback of this approach is that it couples the process of dimesioality reductio with mea subtractio although the two will be show to be idepedet i our derivatio. By takig the derivative of the error fuctio with respect to z k, they also show that z k = W T x k. Hece the approximatio they are seekig is ˆx k = WW T x k + W W T µ. (0)

6 202 A. A. Mirada et al. The secod drawback of their approach is the requiremet of yet aother costraied miimizatio of the error fuctio before they reach the solutio for the optimal colums of W. 5 Methods of PCA We have discussed the eed for a ew derivatio of PCA by (i) explaiig the lack of proper justificatio i the literature for subtractig the mea i a miimum mea square error framework, (ii) remidig its chroic ecessity for the beefit of may applicatios i Sect. 2, ad (iii) reviewig a recet attempt to solve this problem i Sect. 4. Our derivatios of the solutio for the problem i (4) are due to two simple forms of the error fuctio J q of () whichwe state as follows: Form : J q ( ˆx k ) = Form 2 : J q ( ˆx k ) = Tr ( ) T ( ) xk ˆx k xk ˆx k ( ) ( )( ) T xk ˆx k xk ˆx k () (2) We aalyze Form i () i Sect. 6 to arrive at a simplified J q which is exactly the same as we get by followig a differet method of aalyzig Form 2 i (2) i Sect. 7. These two methods pursue differet paths towards the commo error fuctio, viz., the first usig straightforward expasio of the terms i J q ad the secod usig the property of matrix trace. The commo form of J q is subsequetly treated i Sect. 8 to reveal the rest of the solutio to our origial problem. 6 Aalysis of Form of Error Fuctio Usig (7), the error fuctio J q of Form i () ca be developed as J q (B, θ) = x k T BT B x k. (3) The property that B = I WW T is idempotet ad symmetric, i.e., B = B 2 = B T, (4) or B is simply a orthogoal projector, may be used to reduce J q further as Expadig J q above usig x k = x k θ gives J q (B, θ) = x k T B x k. (5) J q (B, θ) = [ ] xk T Bx k 2θ T Bx k + θ T Bθ (6)

7 New Routes from Miimal Approximatio Error to Pricipal Compoets 203 I order to get the θ which miimizes J q, we fid the partial derivative J q / θ = 2B [ x k θ ] ad settig it to zero results i θ = x k = µ, (7) which is as simple as regardig the sample mea of the data poits as the ew origi. Heceforth, we ca assume that x k is the data poit x k from which the sample mea has bee subtracted. 6. Simplifyig the Error Fuctio We may aalyze the error fuctio i (5) as follows: J q (W) = = = x T k ( I WW T ) x k x k T x k x k T x k Tr We have the sample covariace matrix S = x k T WWT x k (W T [ ] ) x k x k T W. x k x k T θ=µ (8) so that the term x T k x k θ=µ equals Tr(S), ad we ca write ( ) J q (W) = Tr(S) Tr W T SW. (9) 7 Aalysis of Form 2 of Error Fuctio We ow aalyze the Form 2 of the error fuctio J q by substitutig (7) i(2)as 7. Fidig θ ( J q (B, θ) = Tr B [ ] ) x k x k T B T. (20) As i the previous sectio, we deote the sample mea ad sample covariace matrix by µ ad S, respectively, ad we may develop the term i (20):

8 204 A. A. Mirada et al. x k x k T = = (x k θ)(x k θ) T [x ] k xk T x kθ T θxk T + θθt = S + µµ T µθ T θµ T + θθ T, (2) where we have used the sample autocorrelatio matrix [4]giveby x k xk T = S+µµT. We get J q (B) = Tr ( B ( S + µµ T µθ T θµ T + θθ T ) B T ) upo substitutig (2)i(20). Usig (4) ad the cyclic permutatio property of trace of matrix products 2 we get ( ( )) J q (B) = Tr B S + µµ T µθ T θµ T + θθ T (22) ad usig the property of the derivative of trace 3 ad the chai rule of the derivatives, 4 we fid that J q / θ = 2B ( µ + θ) which whe equated to zero results i leadig to the same solutio of Form i (7). 7.2 Simplifyig the Error Fuctio θ = µ (23) Havig foud θ, we ca substitute it i (22)togetJ q (B) = Tr (BS). O substitutio for B i terms of W, we may write J q (W) = Tr (S) Tr ( WW T S ). Utilizig the cyclic permutatio property of matrix trace agai, we get J q (W) = Tr (S) Tr ( W T SW ). (24) 8 Optimal Basis ad Miimum Error Note that we have arrived at the same set of equatios i both (9) ad(24) offormad Form 2, respectively, whereby substitutig W as defied i (5) i either of them gives J q (e i ) = Tr(S) 8. Relatio of Variace to Optimal Basis q ei T Se i. (25) Let us ow fid the variace λ i of the data projected o the basis vector e i. It is the average of the square of the differece betwee projectios e T i x k of the data poits ad the projectio i= 2 Tr ( (ϒ ) ( = ))/ Tr ( ϒ ) = Tr ( ϒ). 3 Tr T =. [ ( )] 4 ( )/ u = ( )/ uv T v.

9 New Routes from Miimal Approximatio Error to Pricipal Compoets 205 e T i µ of the sample mea, i.e., λ i = = = e T i ) 2 (ei T x k ei T µ ) T (ei T x k ei )(e T µ i T x k ei T µ [ ] (x k µ)(x k µ) T e i = ei T Se i. (26) Thus, the term q i= et i Se i i (25) gives the portio of the total variace Tr (S) retaied alog the directios of orthoormal e i. Hece, we are lookig for vectors e i of the form λ i = ei T (Se i ), which is satisfied if Se i = λ i e i. Such a relatio implies (e i,λ i ) form a eige-pair of S. Note that sice there is o uique basis for ay otrivial vector space, ay basis that spas the q-dimesioal space geerated by the eigevectors of S are solutios to e i too. I (25), sice argmi e i J q = argmax e i q ei T Se i, (27) the vectors e i have to be the eigevectors correspodig to the q largest ( pricipal ) eigevalues of S. This is the classical result of the PCA. i= 8.2 Relatio of Variace to Miimum Approximatio Error It follows from (26) that the term q i= et i Se i = q i= λ i of (25)isthesumoftheq pricipal eigevalues of S; this is the maximum variace that could be retaied upo approximatio usig ay q basis vectors. Also, Tr (S) = r i= λ i,r = rak (S) is the total variace i the data. Substitutig these i J q i (25) gives the differece of the total variace ad the maximum retaied variace; the result is the miimum of the elimiated variace. Hece, for λ i λ j,j >i, the miimum mea square approximatio error ca be expressed as J q = r λ i i= }{{} total variace q i= λ i }{{} retaied variace = r i=q+ λ i. }{{} elimiated variace 8.3 Compariso of the Reviewed Solutio with the Preset Work (28) I order to compare the solutio of [] reviewed i Sect. 4, let us first write the approximatio i (6)as ˆx k = WW T x k + Bθ. We kow from (7)ad(23)thatθ = µ ad, hece, ˆx k = WW T x k + Bµ. (29) If W W T = B, we have the approximatio accordig to [] i(0) ofsect.4 equivalet to the approximatio i (29). While the drawbacks of (6) highlighted i Sect. 4 exist, let us outlie the differece i these two approaches: we have demostrated i the proposed framework that the ew origi θ R p of the low dimesioal coordiate system should be the mea µ R p so that the

10 206 A. A. Mirada et al. error of the approximatio is reduced. But [] ecessitates a orthogoal projectio of certai data-idepedet compoets b R p q to µ R p to achieve the same objective. The framework preseted i this letter has show that such a dimesioality reductio coupled with mea subtractio is uecessary for derivig PCA. 8.4 Populatio PCA For populatio PCA [0,4], where the samples that form the data are assumed to be realizatios of a radom variable, we have made it easy for the reader to follow our aalysis by just replacig all occurreces of E, the expectatio operator; ad bold faces for radom variables as i x k x, ˆx k ˆx, ad x k x. 9Coclusio Motivated by the eed to justify the heuristics of pre-aalysis mea ceterig i PCA ad related questios, we have demostrated through two distict methods that the mea subtractio becomes part of the solutio of the stadard PCA problem i a approximatio error miimizatio framework. We believe that the framework, i which we have compared a recet solutio with ours, is more effective i justifyig mea subtractio i PCA. Also, we have show that the framework is comprehesive i the sese that the two outcomes of optimal dimesioality reductio, viz. approximatio error miimizatio ad retaied variace maximizatio, are attaied here simultaeously. Ackowledgmets This work was fuded by the Project TANIA (WALEO II) of the Walloo Regio, Belgium. The authors thak their colleague Olivier Caele for his appreciable commets. Thaks are also due to Dr. P.P. Mohalal of ISRO Iertial Systems Uit, Idia for his valuable isights. The authors are very grateful to the Editor-i-Chief ad three aoymous reviewers for their excellet suggestios o a earlier versio of this letter. Refereces. Bishop CM (2006) Patter recogitio ad machie learig. Iformatio sciece ad statistics. Spriger, New York 2. Diamataras KI, Kug SY (996) Pricipal compoet eural etworks: theory ad applicatios. Joh wiley, NewYork 3. Duda RO, Hart PE, Stork DG (200) Patter classificatio. 2d ed. Wiley Itersciece, New York 4. Fukuaga K (990) Itroductio to statistical patter recogitio. Computer sciece ad scietific computig,. 2d ed. Academic Press, Sa Diego 5. Fukuaga K, Kootz WLG (970) Applicatio of the Karhue Loeve expasio to feature selectio ad orderig. IEEE Trasac Comput C-9(4): Harsayi JC, Chag C-I (994) Hyperspectral image classificatio ad dimesioality reductio: a orthogoal subspace projectio approach. IEEE Trasac Geosci Remote Ses 32(4): Hotellig H (933) Aalysis of a complex of statistical variables ito pricipal compoets. J Educ Psychol 24: Huo X, Elad M, Flesia AG, Muise B, Stafill R, Mahalaobis A et al (2003) Optimal reduced-rak quadratic classifiers usig the Fukuaga Kootz trasform with applicatios to automated target recogitio. Proc SPIE 5094: Hyvarie A, Karhue J, Oja E (200) Idepedet compoet aalysis, vol 27 of adaptive ad learig systems for sigal processig, commuicatios ad cotrol. Wiley-Itersciece, New York 0. Johso RA, Wicher DW (992) Applied multivariate statistical aalysis, 3rd ed. Pretice-Hall, Ic., Upper Saddle River. Jolliffe IT (2002) Pricipal compoet aalysis, 2d ed. Spriger, New York

11 New Routes from Miimal Approximatio Error to Pricipal Compoets Mahaalobis A, Muise RR, Stafill SR, Va Nevel A (2004) Desig ad applicatio of quadratic correlatio filters for target detectio. IEEE Trasac Aerosp Electro Syst 40(3): Ma ME, Bradley RS, Hughes MK (998) Global-scale temperature patters ad climate forcig over the past six ceturies. Nature 392: Mardia K, Ket J, Bibby J (979) Multivariate aalysis. Academic Press, Lodo 5. McItyre S, McKitrick R (2005) Reply to commet by Huybers o hockey sticks, pricipal compoets, ad spurious sigificace. Geophys Res Lett 32:L Mirada AA, Whela PF (2005) Fukuaga Kootz trasform for small sample size problems. I: Proceedigs of the IEE Irish sigals ad systems coferece, pp 56 6, Dubli 7. Noy-Meir I (973) Data trasformatios i ecological ordiatio: I. some advatages of o-ceterig. J Ecol 6(2): Pearso K (90) O lies ad plaes of closest fit to systems of poits i space. Philos Mag 2: Plett GL, Doi T, Torrieri D (997) Mie detectio usig scatterig parameters ad a artificial eural etwork. IEEE Trasac Neural Netw 8(6): Ripley BD (996) Patter recogitio ad eural etworks. Cambridge Uiversity Press, Cambridge 2. Va Huffel S (ed) (997) Recet advaces i total least squares techiques ad errors-i-variables modelig. SIAM, Philadelphia

New Routes from Minimal Approximation Error to Principal Components

New Routes from Minimal Approximation Error to Principal Components New Routes from Miimal Approximatio Error to Pricipal Compoets Abhilash Alexader Mirada, Ya-Aël Le Borge, Gialuca Botempi Machie Learig Group, Départemet d Iformatique Uiversité Libre de Bruxelles, Boulevard

More information

Machine Learning for Data Science (CS 4786)

Machine Learning for Data Science (CS 4786) Machie Learig for Data Sciece CS 4786) Lecture & 3: Pricipal Compoet Aalysis The text i black outlies high level ideas. The text i blue provides simple mathematical details to derive or get to the algorithm

More information

Linear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d

Linear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d Liear regressio Daiel Hsu (COMS 477) Maximum likelihood estimatio Oe of the simplest liear regressio models is the followig: (X, Y ),..., (X, Y ), (X, Y ) are iid radom pairs takig values i R d R, ad Y

More information

Machine Learning for Data Science (CS 4786)

Machine Learning for Data Science (CS 4786) Machie Learig for Data Sciece CS 4786) Lecture 9: Pricipal Compoet Aalysis The text i black outlies mai ideas to retai from the lecture. The text i blue give a deeper uderstadig of how we derive or get

More information

Linear Regression Demystified

Linear Regression Demystified Liear Regressio Demystified Liear regressio is a importat subject i statistics. I elemetary statistics courses, formulae related to liear regressio are ofte stated without derivatio. This ote iteds to

More information

6. Kalman filter implementation for linear algebraic equations. Karhunen-Loeve decomposition

6. Kalman filter implementation for linear algebraic equations. Karhunen-Loeve decomposition 6. Kalma filter implemetatio for liear algebraic equatios. Karhue-Loeve decompositio 6.1. Solvable liear algebraic systems. Probabilistic iterpretatio. Let A be a quadratic matrix (ot obligatory osigular.

More information

Session 5. (1) Principal component analysis and Karhunen-Loève transformation

Session 5. (1) Principal component analysis and Karhunen-Loève transformation 200 Autum semester Patter Iformatio Processig Topic 2 Image compressio by orthogoal trasformatio Sessio 5 () Pricipal compoet aalysis ad Karhue-Loève trasformatio Topic 2 of this course explais the image

More information

Discrete Orthogonal Moment Features Using Chebyshev Polynomials

Discrete Orthogonal Moment Features Using Chebyshev Polynomials Discrete Orthogoal Momet Features Usig Chebyshev Polyomials R. Mukuda, 1 S.H.Og ad P.A. Lee 3 1 Faculty of Iformatio Sciece ad Techology, Multimedia Uiversity 75450 Malacca, Malaysia. Istitute of Mathematical

More information

Similarity Solutions to Unsteady Pseudoplastic. Flow Near a Moving Wall

Similarity Solutions to Unsteady Pseudoplastic. Flow Near a Moving Wall Iteratioal Mathematical Forum, Vol. 9, 04, o. 3, 465-475 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/0.988/imf.04.48 Similarity Solutios to Usteady Pseudoplastic Flow Near a Movig Wall W. Robi Egieerig

More information

The Method of Least Squares. To understand least squares fitting of data.

The Method of Least Squares. To understand least squares fitting of data. The Method of Least Squares KEY WORDS Curve fittig, least square GOAL To uderstad least squares fittig of data To uderstad the least squares solutio of icosistet systems of liear equatios 1 Motivatio Curve

More information

ECON 3150/4150, Spring term Lecture 3

ECON 3150/4150, Spring term Lecture 3 Itroductio Fidig the best fit by regressio Residuals ad R-sq Regressio ad causality Summary ad ext step ECON 3150/4150, Sprig term 2014. Lecture 3 Ragar Nymoe Uiversity of Oslo 21 Jauary 2014 1 / 30 Itroductio

More information

A NOTE ON THE TOTAL LEAST SQUARES FIT TO COPLANAR POINTS

A NOTE ON THE TOTAL LEAST SQUARES FIT TO COPLANAR POINTS A NOTE ON THE TOTAL LEAST SQUARES FIT TO COPLANAR POINTS STEVEN L. LEE Abstract. The Total Least Squares (TLS) fit to the poits (x,y ), =1,,, miimizes the sum of the squares of the perpedicular distaces

More information

Geometry of LS. LECTURE 3 GEOMETRY OF LS, PROPERTIES OF σ 2, PARTITIONED REGRESSION, GOODNESS OF FIT

Geometry of LS. LECTURE 3 GEOMETRY OF LS, PROPERTIES OF σ 2, PARTITIONED REGRESSION, GOODNESS OF FIT OCTOBER 7, 2016 LECTURE 3 GEOMETRY OF LS, PROPERTIES OF σ 2, PARTITIONED REGRESSION, GOODNESS OF FIT Geometry of LS We ca thik of y ad the colums of X as members of the -dimesioal Euclidea space R Oe ca

More information

Definitions and Theorems. where x are the decision variables. c, b, and a are constant coefficients.

Definitions and Theorems. where x are the decision variables. c, b, and a are constant coefficients. Defiitios ad Theorems Remember the scalar form of the liear programmig problem, Miimize, Subject to, f(x) = c i x i a 1i x i = b 1 a mi x i = b m x i 0 i = 1,2,, where x are the decisio variables. c, b,

More information

Support vector machine revisited

Support vector machine revisited 6.867 Machie learig, lecture 8 (Jaakkola) 1 Lecture topics: Support vector machie ad kerels Kerel optimizatio, selectio Support vector machie revisited Our task here is to first tur the support vector

More information

5.1 Review of Singular Value Decomposition (SVD)

5.1 Review of Singular Value Decomposition (SVD) MGMT 69000: Topics i High-dimesioal Data Aalysis Falll 06 Lecture 5: Spectral Clusterig: Overview (cotd) ad Aalysis Lecturer: Jiamig Xu Scribe: Adarsh Barik, Taotao He, September 3, 06 Outlie Review of

More information

Outline. Linear regression. Regularization functions. Polynomial curve fitting. Stochastic gradient descent for regression. MLE for regression

Outline. Linear regression. Regularization functions. Polynomial curve fitting. Stochastic gradient descent for regression. MLE for regression REGRESSION 1 Outlie Liear regressio Regularizatio fuctios Polyomial curve fittig Stochastic gradiet descet for regressio MLE for regressio Step-wise forward regressio Regressio methods Statistical techiques

More information

Apply change-of-basis formula to rewrite x as a linear combination of eigenvectors v j.

Apply change-of-basis formula to rewrite x as a linear combination of eigenvectors v j. Eigevalue-Eigevector Istructor: Nam Su Wag eigemcd Ay vector i real Euclidea space of dimesio ca be uiquely epressed as a liear combiatio of liearly idepedet vectors (ie, basis) g j, j,,, α g α g α g α

More information

State Space Representation

State Space Representation Optimal Cotrol, Guidace ad Estimatio Lecture 2 Overview of SS Approach ad Matrix heory Prof. Radhakat Padhi Dept. of Aerospace Egieerig Idia Istitute of Sciece - Bagalore State Space Represetatio Prof.

More information

Algebra of Least Squares

Algebra of Least Squares October 19, 2018 Algebra of Least Squares Geometry of Least Squares Recall that out data is like a table [Y X] where Y collects observatios o the depedet variable Y ad X collects observatios o the k-dimesioal

More information

Topics in Eigen-analysis

Topics in Eigen-analysis Topics i Eige-aalysis Li Zajiag 28 July 2014 Cotets 1 Termiology... 2 2 Some Basic Properties ad Results... 2 3 Eige-properties of Hermitia Matrices... 5 3.1 Basic Theorems... 5 3.2 Quadratic Forms & Noegative

More information

Introduction to Machine Learning DIS10

Introduction to Machine Learning DIS10 CS 189 Fall 017 Itroductio to Machie Learig DIS10 1 Fu with Lagrage Multipliers (a) Miimize the fuctio such that f (x,y) = x + y x + y = 3. Solutio: The Lagragia is: L(x,y,λ) = x + y + λ(x + y 3) Takig

More information

BIOINF 585: Machine Learning for Systems Biology & Clinical Informatics

BIOINF 585: Machine Learning for Systems Biology & Clinical Informatics BIOINF 585: Machie Learig for Systems Biology & Cliical Iformatics Lecture 14: Dimesio Reductio Jie Wag Departmet of Computatioal Medicie & Bioiformatics Uiversity of Michiga 1 Outlie What is feature reductio?

More information

10/2/ , 5.9, Jacob Hays Amit Pillay James DeFelice

10/2/ , 5.9, Jacob Hays Amit Pillay James DeFelice 0//008 Liear Discrimiat Fuctios Jacob Hays Amit Pillay James DeFelice 5.8, 5.9, 5. Miimum Squared Error Previous methods oly worked o liear separable cases, by lookig at misclassified samples to correct

More information

Summary: CORRELATION & LINEAR REGRESSION. GC. Students are advised to refer to lecture notes for the GC operations to obtain scatter diagram.

Summary: CORRELATION & LINEAR REGRESSION. GC. Students are advised to refer to lecture notes for the GC operations to obtain scatter diagram. Key Cocepts: 1) Sketchig of scatter diagram The scatter diagram of bivariate (i.e. cotaiig two variables) data ca be easily obtaied usig GC. Studets are advised to refer to lecture otes for the GC operatios

More information

A Lattice Green Function Introduction. Abstract

A Lattice Green Function Introduction. Abstract August 5, 25 A Lattice Gree Fuctio Itroductio Stefa Hollos Exstrom Laboratories LLC, 662 Nelso Park Dr, Logmot, Colorado 853, USA Abstract We preset a itroductio to lattice Gree fuctios. Electroic address:

More information

Cov(aX, cy ) Var(X) Var(Y ) It is completely invariant to affine transformations: for any a, b, c, d R, ρ(ax + b, cy + d) = a.s. X i. as n.

Cov(aX, cy ) Var(X) Var(Y ) It is completely invariant to affine transformations: for any a, b, c, d R, ρ(ax + b, cy + d) = a.s. X i. as n. CS 189 Itroductio to Machie Learig Sprig 218 Note 11 1 Caoical Correlatio Aalysis The Pearso Correlatio Coefficiet ρ(x, Y ) is a way to measure how liearly related (i other words, how well a liear model

More information

6.3 Testing Series With Positive Terms

6.3 Testing Series With Positive Terms 6.3. TESTING SERIES WITH POSITIVE TERMS 307 6.3 Testig Series With Positive Terms 6.3. Review of what is kow up to ow I theory, testig a series a i for covergece amouts to fidig the i= sequece of partial

More information

U8L1: Sec Equations of Lines in R 2

U8L1: Sec Equations of Lines in R 2 MCVU U8L: Sec. 8.9. Equatios of Lies i R Review of Equatios of a Straight Lie (-D) Cosider the lie passig through A (-,) with slope, as show i the diagram below. I poit slope form, the equatio of the lie

More information

1 Duality revisited. AM 221: Advanced Optimization Spring 2016

1 Duality revisited. AM 221: Advanced Optimization Spring 2016 AM 22: Advaced Optimizatio Sprig 206 Prof. Yaro Siger Sectio 7 Wedesday, Mar. 9th Duality revisited I this sectio, we will give a slightly differet perspective o duality. optimizatio program: f(x) x R

More information

4. Hypothesis testing (Hotelling s T 2 -statistic)

4. Hypothesis testing (Hotelling s T 2 -statistic) 4. Hypothesis testig (Hotellig s T -statistic) Cosider the test of hypothesis H 0 : = 0 H A = 6= 0 4. The Uio-Itersectio Priciple W accept the hypothesis H 0 as valid if ad oly if H 0 (a) : a T = a T 0

More information

A Risk Comparison of Ordinary Least Squares vs Ridge Regression

A Risk Comparison of Ordinary Least Squares vs Ridge Regression Joural of Machie Learig Research 14 (2013) 1505-1511 Submitted 5/12; Revised 3/13; Published 6/13 A Risk Compariso of Ordiary Least Squares vs Ridge Regressio Paramveer S. Dhillo Departmet of Computer

More information

Linear Classifiers III

Linear Classifiers III Uiversität Potsdam Istitut für Iformatik Lehrstuhl Maschielles Lere Liear Classifiers III Blaie Nelso, Tobias Scheffer Cotets Classificatio Problem Bayesia Classifier Decisio Liear Classifiers, MAP Models

More information

ECE 901 Lecture 12: Complexity Regularization and the Squared Loss

ECE 901 Lecture 12: Complexity Regularization and the Squared Loss ECE 90 Lecture : Complexity Regularizatio ad the Squared Loss R. Nowak 5/7/009 I the previous lectures we made use of the Cheroff/Hoeffdig bouds for our aalysis of classifier errors. Hoeffdig s iequality

More information

3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense,

3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense, 3. Z Trasform Referece: Etire Chapter 3 of text. Recall that the Fourier trasform (FT) of a DT sigal x [ ] is ω ( ) [ ] X e = j jω k = xe I order for the FT to exist i the fiite magitude sese, S = x [

More information

Chapter Vectors

Chapter Vectors Chapter 4. Vectors fter readig this chapter you should be able to:. defie a vector. add ad subtract vectors. fid liear combiatios of vectors ad their relatioship to a set of equatios 4. explai what it

More information

Symmetric Matrices and Quadratic Forms

Symmetric Matrices and Quadratic Forms 7 Symmetric Matrices ad Quadratic Forms 7.1 DIAGONALIZAION OF SYMMERIC MARICES SYMMERIC MARIX A symmetric matrix is a matrix A such that. A = A Such a matrix is ecessarily square. Its mai diagoal etries

More information

September 2012 C1 Note. C1 Notes (Edexcel) Copyright - For AS, A2 notes and IGCSE / GCSE worksheets 1

September 2012 C1 Note. C1 Notes (Edexcel) Copyright   - For AS, A2 notes and IGCSE / GCSE worksheets 1 September 0 s (Edecel) Copyright www.pgmaths.co.uk - For AS, A otes ad IGCSE / GCSE worksheets September 0 Copyright www.pgmaths.co.uk - For AS, A otes ad IGCSE / GCSE worksheets September 0 Copyright

More information

1 Adiabatic and diabatic representations

1 Adiabatic and diabatic representations 1 Adiabatic ad diabatic represetatios 1.1 Bor-Oppeheimer approximatio The time-idepedet Schrödiger equatio for both electroic ad uclear degrees of freedom is Ĥ Ψ(r, R) = E Ψ(r, R), (1) where the full molecular

More information

Axis Aligned Ellipsoid

Axis Aligned Ellipsoid Machie Learig for Data Sciece CS 4786) Lecture 6,7 & 8: Ellipsoidal Clusterig, Gaussia Mixture Models ad Geeral Mixture Models The text i black outlies high level ideas. The text i blue provides simple

More information

Matrix Representation of Data in Experiment

Matrix Representation of Data in Experiment Matrix Represetatio of Data i Experimet Cosider a very simple model for resposes y ij : y ij i ij, i 1,; j 1,,..., (ote that for simplicity we are assumig the two () groups are of equal sample size ) Y

More information

Infinite Sequences and Series

Infinite Sequences and Series Chapter 6 Ifiite Sequeces ad Series 6.1 Ifiite Sequeces 6.1.1 Elemetary Cocepts Simply speakig, a sequece is a ordered list of umbers writte: {a 1, a 2, a 3,...a, a +1,...} where the elemets a i represet

More information

Vector Quantization: a Limiting Case of EM

Vector Quantization: a Limiting Case of EM . Itroductio & defiitios Assume that you are give a data set X = { x j }, j { 2,,, }, of d -dimesioal vectors. The vector quatizatio (VQ) problem requires that we fid a set of prototype vectors Z = { z

More information

Lecture 7: Density Estimation: k-nearest Neighbor and Basis Approach

Lecture 7: Density Estimation: k-nearest Neighbor and Basis Approach STAT 425: Itroductio to Noparametric Statistics Witer 28 Lecture 7: Desity Estimatio: k-nearest Neighbor ad Basis Approach Istructor: Ye-Chi Che Referece: Sectio 8.4 of All of Noparametric Statistics.

More information

4.1 Sigma Notation and Riemann Sums

4.1 Sigma Notation and Riemann Sums 0 the itegral. Sigma Notatio ad Riema Sums Oe strategy for calculatig the area of a regio is to cut the regio ito simple shapes, calculate the area of each simple shape, ad the add these smaller areas

More information

Chandrasekhar Type Algorithms. for the Riccati Equation of Lainiotis Filter

Chandrasekhar Type Algorithms. for the Riccati Equation of Lainiotis Filter Cotemporary Egieerig Scieces, Vol. 3, 00, o. 4, 9-00 Chadrasekhar ype Algorithms for the Riccati Equatio of Laiiotis Filter Nicholas Assimakis Departmet of Electroics echological Educatioal Istitute of

More information

Variable selection in principal components analysis of qualitative data using the accelerated ALS algorithm

Variable selection in principal components analysis of qualitative data using the accelerated ALS algorithm Variable selectio i pricipal compoets aalysis of qualitative data usig the accelerated ALS algorithm Masahiro Kuroda Yuichi Mori Masaya Iizuka Michio Sakakihara (Okayama Uiversity of Sciece) (Okayama Uiversity

More information

Factor Analysis. Lecture 10: Factor Analysis and Principal Component Analysis. Sam Roweis

Factor Analysis. Lecture 10: Factor Analysis and Principal Component Analysis. Sam Roweis Lecture 10: Factor Aalysis ad Pricipal Compoet Aalysis Sam Roweis February 9, 2004 Whe we assume that the subspace is liear ad that the uderlyig latet variable has a Gaussia distributio we get a model

More information

Achieving Stationary Distributions in Markov Chains. Monday, November 17, 2008 Rice University

Achieving Stationary Distributions in Markov Chains. Monday, November 17, 2008 Rice University Istructor: Achievig Statioary Distributios i Markov Chais Moday, November 1, 008 Rice Uiversity Dr. Volka Cevher STAT 1 / ELEC 9: Graphical Models Scribe: Rya E. Guerra, Tahira N. Saleem, Terrace D. Savitsky

More information

ALGEBRAIC GEOMETRY COURSE NOTES, LECTURE 5: SINGULARITIES.

ALGEBRAIC GEOMETRY COURSE NOTES, LECTURE 5: SINGULARITIES. ALGEBRAIC GEOMETRY COURSE NOTES, LECTURE 5: SINGULARITIES. ANDREW SALCH 1. The Jacobia criterio for osigularity. You have probably oticed by ow that some poits o varieties are smooth i a sese somethig

More information

Correlation Regression

Correlation Regression Correlatio Regressio While correlatio methods measure the stregth of a liear relatioship betwee two variables, we might wish to go a little further: How much does oe variable chage for a give chage i aother

More information

Dimensionality Reduction vs. Clustering

Dimensionality Reduction vs. Clustering Dimesioality Reductio vs. Clusterig Lecture 9: Cotiuous Latet Variable Models Sam Roweis Traiig such factor models (e.g. FA, PCA, ICA) is called dimesioality reductio. You ca thik of this as (o)liear regressio

More information

We are mainly going to be concerned with power series in x, such as. (x)} converges - that is, lims N n

We are mainly going to be concerned with power series in x, such as. (x)} converges - that is, lims N n Review of Power Series, Power Series Solutios A power series i x - a is a ifiite series of the form c (x a) =c +c (x a)+(x a) +... We also call this a power series cetered at a. Ex. (x+) is cetered at

More information

Estimation of Population Mean Using Co-Efficient of Variation and Median of an Auxiliary Variable

Estimation of Population Mean Using Co-Efficient of Variation and Median of an Auxiliary Variable Iteratioal Joural of Probability ad Statistics 01, 1(4: 111-118 DOI: 10.593/j.ijps.010104.04 Estimatio of Populatio Mea Usig Co-Efficiet of Variatio ad Media of a Auxiliary Variable J. Subramai *, G. Kumarapadiya

More information

On Nonsingularity of Saddle Point Matrices. with Vectors of Ones

On Nonsingularity of Saddle Point Matrices. with Vectors of Ones Iteratioal Joural of Algebra, Vol. 2, 2008, o. 4, 197-204 O Nosigularity of Saddle Poit Matrices with Vectors of Oes Tadeusz Ostrowski Istitute of Maagemet The State Vocatioal Uiversity -400 Gorzów, Polad

More information

Lecture 12: February 28

Lecture 12: February 28 10-716: Advaced Machie Learig Sprig 2019 Lecture 12: February 28 Lecturer: Pradeep Ravikumar Scribes: Jacob Tyo, Rishub Jai, Ojash Neopae Note: LaTeX template courtesy of UC Berkeley EECS dept. Disclaimer:

More information

Lecture 8: October 20, Applications of SVD: least squares approximation

Lecture 8: October 20, Applications of SVD: least squares approximation Mathematical Toolkit Autum 2016 Lecturer: Madhur Tulsiai Lecture 8: October 20, 2016 1 Applicatios of SVD: least squares approximatio We discuss aother applicatio of sigular value decompositio (SVD) of

More information

Riesz-Fischer Sequences and Lower Frame Bounds

Riesz-Fischer Sequences and Lower Frame Bounds Zeitschrift für Aalysis ud ihre Aweduge Joural for Aalysis ad its Applicatios Volume 1 (00), No., 305 314 Riesz-Fischer Sequeces ad Lower Frame Bouds P. Casazza, O. Christese, S. Li ad A. Lider Abstract.

More information

Machine Learning for Data Science (CS4786) Lecture 9

Machine Learning for Data Science (CS4786) Lecture 9 Machie Learig for Data Sciece (CS4786) Lecture 9 Pricipal Compoet Aalysis Course Webpage : http://www.cs.corell.eu/courses/cs4786/207fa/ DIM REDUCTION: LINEAR TRANSFORMATION x > y > Pick a low imesioal

More information

(3) If you replace row i of A by its sum with a multiple of another row, then the determinant is unchanged! Expand across the i th row:

(3) If you replace row i of A by its sum with a multiple of another row, then the determinant is unchanged! Expand across the i th row: Math 5-4 Tue Feb 4 Cotiue with sectio 36 Determiats The effective way to compute determiats for larger-sized matrices without lots of zeroes is to ot use the defiitio, but rather to use the followig facts,

More information

Discrete-Time Systems, LTI Systems, and Discrete-Time Convolution

Discrete-Time Systems, LTI Systems, and Discrete-Time Convolution EEL5: Discrete-Time Sigals ad Systems. Itroductio I this set of otes, we begi our mathematical treatmet of discrete-time s. As show i Figure, a discrete-time operates or trasforms some iput sequece x [

More information

Stochastic Matrices in a Finite Field

Stochastic Matrices in a Finite Field Stochastic Matrices i a Fiite Field Abstract: I this project we will explore the properties of stochastic matrices i both the real ad the fiite fields. We first explore what properties 2 2 stochastic matrices

More information

LECTURE 8: ORTHOGONALITY (CHAPTER 5 IN THE BOOK)

LECTURE 8: ORTHOGONALITY (CHAPTER 5 IN THE BOOK) LECTURE 8: ORTHOGONALITY (CHAPTER 5 IN THE BOOK) Everythig marked by is ot required by the course syllabus I this lecture, all vector spaces is over the real umber R. All vectors i R is viewed as a colum

More information

Inverse Matrix. A meaning that matrix B is an inverse of matrix A.

Inverse Matrix. A meaning that matrix B is an inverse of matrix A. Iverse Matrix Two square matrices A ad B of dimesios are called iverses to oe aother if the followig holds, AB BA I (11) The otio is dual but we ofte write 1 B A meaig that matrix B is a iverse of matrix

More information

Singular value decomposition. Mathématiques appliquées (MATH0504-1) B. Dewals, Ch. Geuzaine

Singular value decomposition. Mathématiques appliquées (MATH0504-1) B. Dewals, Ch. Geuzaine Lecture 11 Sigular value decompositio Mathématiques appliquées (MATH0504-1) B. Dewals, Ch. Geuzaie V1.2 07/12/2018 1 Sigular value decompositio (SVD) at a glace Motivatio: the image of the uit sphere S

More information

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 3

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 3 Machie Learig Theory Tübige Uiversity, WS 06/07 Lecture 3 Tolstikhi Ilya Abstract I this lecture we will prove the VC-boud, which provides a high-probability excess risk boud for the ERM algorithm whe

More information

c 2006 Society for Industrial and Applied Mathematics

c 2006 Society for Industrial and Applied Mathematics SIAM J. MATRIX ANAL. APPL. Vol. 7, No. 3, pp. 851 860 c 006 Society for Idustrial ad Applied Mathematics EXTREMAL EIGENVALUES OF REAL SYMMETRIC MATRICES WITH ENTRIES IN AN INTERVAL XINGZHI ZHAN Abstract.

More information

Lecture 22: Review for Exam 2. 1 Basic Model Assumptions (without Gaussian Noise)

Lecture 22: Review for Exam 2. 1 Basic Model Assumptions (without Gaussian Noise) Lecture 22: Review for Exam 2 Basic Model Assumptios (without Gaussia Noise) We model oe cotiuous respose variable Y, as a liear fuctio of p umerical predictors, plus oise: Y = β 0 + β X +... β p X p +

More information

THE KALMAN FILTER RAUL ROJAS

THE KALMAN FILTER RAUL ROJAS THE KALMAN FILTER RAUL ROJAS Abstract. This paper provides a getle itroductio to the Kalma filter, a umerical method that ca be used for sesor fusio or for calculatio of trajectories. First, we cosider

More information

Machine Learning Regression I Hamid R. Rabiee [Slides are based on Bishop Book] Spring

Machine Learning Regression I Hamid R. Rabiee [Slides are based on Bishop Book] Spring Machie Learig Regressio I Hamid R. Rabiee [Slides are based o Bishop Book] Sprig 015 http://ce.sharif.edu/courses/93-94//ce717-1 Liear Regressio Liear regressio: ivolves a respose variable ad a sigle predictor

More information

Some New Iterative Methods for Solving Nonlinear Equations

Some New Iterative Methods for Solving Nonlinear Equations World Applied Scieces Joural 0 (6): 870-874, 01 ISSN 1818-495 IDOSI Publicatios, 01 DOI: 10.589/idosi.wasj.01.0.06.830 Some New Iterative Methods for Solvig Noliear Equatios Muhammad Aslam Noor, Khalida

More information

Preponderantly increasing/decreasing data in regression analysis

Preponderantly increasing/decreasing data in regression analysis Croatia Operatioal Research Review 269 CRORR 7(2016), 269 276 Prepoderatly icreasig/decreasig data i regressio aalysis Darija Marković 1, 1 Departmet of Mathematics, J. J. Strossmayer Uiversity of Osijek,

More information

CHAPTER 5. Theory and Solution Using Matrix Techniques

CHAPTER 5. Theory and Solution Using Matrix Techniques A SERIES OF CLASS NOTES FOR 2005-2006 TO INTRODUCE LINEAR AND NONLINEAR PROBLEMS TO ENGINEERS, SCIENTISTS, AND APPLIED MATHEMATICIANS DE CLASS NOTES 3 A COLLECTION OF HANDOUTS ON SYSTEMS OF ORDINARY DIFFERENTIAL

More information

Physics 324, Fall Dirac Notation. These notes were produced by David Kaplan for Phys. 324 in Autumn 2001.

Physics 324, Fall Dirac Notation. These notes were produced by David Kaplan for Phys. 324 in Autumn 2001. Physics 324, Fall 2002 Dirac Notatio These otes were produced by David Kapla for Phys. 324 i Autum 2001. 1 Vectors 1.1 Ier product Recall from liear algebra: we ca represet a vector V as a colum vector;

More information

Topic 9: Sampling Distributions of Estimators

Topic 9: Sampling Distributions of Estimators Topic 9: Samplig Distributios of Estimators Course 003, 2016 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be

More information

Beurling Integers: Part 2

Beurling Integers: Part 2 Beurlig Itegers: Part 2 Isomorphisms Devi Platt July 11, 2015 1 Prime Factorizatio Sequeces I the last article we itroduced the Beurlig geeralized itegers, which ca be represeted as a sequece of real umbers

More information

A class of spectral bounds for Max k-cut

A class of spectral bounds for Max k-cut A class of spectral bouds for Max k-cut Miguel F. Ajos, José Neto December 07 Abstract Let G be a udirected ad edge-weighted simple graph. I this paper we itroduce a class of bouds for the maximum k-cut

More information

An Introduction to Randomized Algorithms

An Introduction to Randomized Algorithms A Itroductio to Radomized Algorithms The focus of this lecture is to study a radomized algorithm for quick sort, aalyze it usig probabilistic recurrece relatios, ad also provide more geeral tools for aalysis

More information

A statistical method to determine sample size to estimate characteristic value of soil parameters

A statistical method to determine sample size to estimate characteristic value of soil parameters A statistical method to determie sample size to estimate characteristic value of soil parameters Y. Hojo, B. Setiawa 2 ad M. Suzuki 3 Abstract Sample size is a importat factor to be cosidered i determiig

More information

Problem Set 4 Due Oct, 12

Problem Set 4 Due Oct, 12 EE226: Radom Processes i Systems Lecturer: Jea C. Walrad Problem Set 4 Due Oct, 12 Fall 06 GSI: Assae Gueye This problem set essetially reviews detectio theory ad hypothesis testig ad some basic otios

More information

REVISION SHEET FP1 (MEI) ALGEBRA. Identities In mathematics, an identity is a statement which is true for all values of the variables it contains.

REVISION SHEET FP1 (MEI) ALGEBRA. Identities In mathematics, an identity is a statement which is true for all values of the variables it contains. The mai ideas are: Idetities REVISION SHEET FP (MEI) ALGEBRA Before the exam you should kow: If a expressio is a idetity the it is true for all values of the variable it cotais The relatioships betwee

More information

Chimica Inorganica 3

Chimica Inorganica 3 himica Iorgaica Irreducible Represetatios ad haracter Tables Rather tha usig geometrical operatios, it is ofte much more coveiet to employ a ew set of group elemets which are matrices ad to make the rule

More information

Clustering. CM226: Machine Learning for Bioinformatics. Fall Sriram Sankararaman Acknowledgments: Fei Sha, Ameet Talwalkar.

Clustering. CM226: Machine Learning for Bioinformatics. Fall Sriram Sankararaman Acknowledgments: Fei Sha, Ameet Talwalkar. Clusterig CM226: Machie Learig for Bioiformatics. Fall 216 Sriram Sakararama Ackowledgmets: Fei Sha, Ameet Talwalkar Clusterig 1 / 42 Admiistratio HW 1 due o Moday. Email/post o CCLE if you have questios.

More information

Statistical Inference Based on Extremum Estimators

Statistical Inference Based on Extremum Estimators T. Rotheberg Fall, 2007 Statistical Iferece Based o Extremum Estimators Itroductio Suppose 0, the true value of a p-dimesioal parameter, is kow to lie i some subset S R p : Ofte we choose to estimate 0

More information

Random Matrices with Blocks of Intermediate Scale Strongly Correlated Band Matrices

Random Matrices with Blocks of Intermediate Scale Strongly Correlated Band Matrices Radom Matrices with Blocks of Itermediate Scale Strogly Correlated Bad Matrices Jiayi Tog Advisor: Dr. Todd Kemp May 30, 07 Departmet of Mathematics Uiversity of Califoria, Sa Diego Cotets Itroductio Notatio

More information

C. Complex Numbers. x 6x + 2 = 0. This equation was known to have three real roots, given by simple combinations of the expressions

C. Complex Numbers. x 6x + 2 = 0. This equation was known to have three real roots, given by simple combinations of the expressions C. Complex Numbers. Complex arithmetic. Most people thik that complex umbers arose from attempts to solve quadratic equatios, but actually it was i coectio with cubic equatios they first appeared. Everyoe

More information

w (1) ˆx w (1) x (1) /ρ and w (2) ˆx w (2) x (2) /ρ.

w (1) ˆx w (1) x (1) /ρ and w (2) ˆx w (2) x (2) /ρ. 2 5. Weighted umber of late jobs 5.1. Release dates ad due dates: maximimizig the weight of o-time jobs Oce we add release dates, miimizig the umber of late jobs becomes a sigificatly harder problem. For

More information

18.01 Calculus Jason Starr Fall 2005

18.01 Calculus Jason Starr Fall 2005 Lecture 18. October 5, 005 Homework. Problem Set 5 Part I: (c). Practice Problems. Course Reader: 3G 1, 3G, 3G 4, 3G 5. 1. Approximatig Riema itegrals. Ofte, there is o simpler expressio for the atiderivative

More information

Chi-Squared Tests Math 6070, Spring 2006

Chi-Squared Tests Math 6070, Spring 2006 Chi-Squared Tests Math 6070, Sprig 2006 Davar Khoshevisa Uiversity of Utah February XXX, 2006 Cotets MLE for Goodess-of Fit 2 2 The Multiomial Distributio 3 3 Applicatio to Goodess-of-Fit 6 3 Testig for

More information

Resampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n.

Resampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n. Jauary 1, 2019 Resamplig Methods Motivatio We have so may estimators with the property θ θ d N 0, σ 2 We ca also write θ a N θ, σ 2 /, where a meas approximately distributed as Oce we have a cosistet estimator

More information

Introduction to Optimization Techniques

Introduction to Optimization Techniques Itroductio to Optimizatio Techiques Basic Cocepts of Aalysis - Real Aalysis, Fuctioal Aalysis 1 Basic Cocepts of Aalysis Liear Vector Spaces Defiitio: A vector space X is a set of elemets called vectors

More information

Soo King Lim Figure 1: Figure 2: Figure 3: Figure 4: Figure 5: Figure 6: Figure 7:

Soo King Lim Figure 1: Figure 2: Figure 3: Figure 4: Figure 5: Figure 6: Figure 7: 0 Multivariate Cotrol Chart 3 Multivariate Normal Distributio 5 Estimatio of the Mea ad Covariace Matrix 6 Hotellig s Cotrol Chart 6 Hotellig s Square 8 Average Value of k Subgroups 0 Example 3 3 Value

More information

CEE 522 Autumn Uncertainty Concepts for Geotechnical Engineering

CEE 522 Autumn Uncertainty Concepts for Geotechnical Engineering CEE 5 Autum 005 Ucertaity Cocepts for Geotechical Egieerig Basic Termiology Set A set is a collectio of (mutually exclusive) objects or evets. The sample space is the (collectively exhaustive) collectio

More information

Recursive Algorithms. Recurrences. Recursive Algorithms Analysis

Recursive Algorithms. Recurrences. Recursive Algorithms Analysis Recursive Algorithms Recurreces Computer Sciece & Egieerig 35: Discrete Mathematics Christopher M Bourke cbourke@cseuledu A recursive algorithm is oe i which objects are defied i terms of other objects

More information

The picture in figure 1.1 helps us to see that the area represents the distance traveled. Figure 1: Area represents distance travelled

The picture in figure 1.1 helps us to see that the area represents the distance traveled. Figure 1: Area represents distance travelled 1 Lecture : Area Area ad distace traveled Approximatig area by rectagles Summatio The area uder a parabola 1.1 Area ad distace Suppose we have the followig iformatio about the velocity of a particle, how

More information

CMSE 820: Math. Foundations of Data Sci.

CMSE 820: Math. Foundations of Data Sci. Lecture 17 8.4 Weighted path graphs Take from [10, Lecture 3] As alluded to at the ed of the previous sectio, we ow aalyze weighted path graphs. To that ed, we prove the followig: Theorem 6 (Fiedler).

More information

Bivariate Sample Statistics Geog 210C Introduction to Spatial Data Analysis. Chris Funk. Lecture 7

Bivariate Sample Statistics Geog 210C Introduction to Spatial Data Analysis. Chris Funk. Lecture 7 Bivariate Sample Statistics Geog 210C Itroductio to Spatial Data Aalysis Chris Fuk Lecture 7 Overview Real statistical applicatio: Remote moitorig of east Africa log rais Lead up to Lab 5-6 Review of bivariate/multivariate

More information

Stochastic Simulation

Stochastic Simulation Stochastic Simulatio 1 Itroductio Readig Assigmet: Read Chapter 1 of text. We shall itroduce may of the key issues to be discussed i this course via a couple of model problems. Model Problem 1 (Jackso

More information

2 Geometric interpretation of complex numbers

2 Geometric interpretation of complex numbers 2 Geometric iterpretatio of complex umbers 2.1 Defiitio I will start fially with a precise defiitio, assumig that such mathematical object as vector space R 2 is well familiar to the studets. Recall that

More information

A Note On The Exponential Of A Matrix Whose Elements Are All 1

A Note On The Exponential Of A Matrix Whose Elements Are All 1 Applied Mathematics E-Notes, 8(208), 92-99 c ISSN 607-250 Available free at mirror sites of http://wwwmaththuedutw/ ame/ A Note O The Expoetial Of A Matrix Whose Elemets Are All Reza Farhadia Received

More information