Performance Analysis and Coding Strategy of ECOC SVMs


 Kevin Gordon
 3 years ago
 Views:
Transcription
1 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04), pp Performance Analyss and Codng Strategy of ECOC SVMs Zhgang Yan, and Yuanxuan Yang, School of Envronmental Scence and Spatal Informatcs, Chna Unversty of Mnng and Technology, Xuzhou, Jangsu, P.R.Chna Jangsu Key Laboratory of Resources & Envronmental Informaton Engneerng, Chna Unversty of Mnng and Technology, Xuzhou, Jangsu, P.R.Chna Correspondng author: Zhgang Yan, Abstract The theoretcal upper bound of generalzaton error for ECOC SVMs s derved based on FatShatterng dmensonalty and coverng number. The factors affectng the generalzaton performance of ECOC SVMs are analyzed. From the analyss, t s beleved that n real classfcaton tass, the performance of ECOC depends on the performance of the classfers correspondng to ts codng columns, whch s rrelevant to the mathematcal characterstcs of the ECOC tself. The essence of ECOC SVMs s how to construct an optmal votng machne consstng of a number of SVMs, how to choose SubSVMs whch have better generalzaton ablty, and how to determne the number of SubSVMs tang part n votng, that s the most mportant ssue. Data sets ncludng Segment are selected for test. All the ECOC code columns are constructed usng an exhaustve technque. A SubSVM s traned for each code column, and the generalzaton ablty of each SubSVM s evaluated by classfcaton ntervals and error rates estmated by cross valdaton. Then, all the ECOC code columns are sorted by the generalzaton performance of SubSVMs. Three categores of ECOC SVMs, ncludng superor, nferor and ordnary categores, are constructed from the sorted ECOC code columns, by usng forward, bacward and orgnal sequences. Expermental results show that the performance of ECOC SVMs whch consst of SubSVMs wth better generalzaton ablty s better and vce versa, whch valdates our vew and ponts out the drecton for mprovng ECOC SVMs. Keywords: ECOC, SVM, Generalzaton Ablty, Code Matrx. Introducton Numerous supervsed learnng algorthms are desgned for twoclass problems, for example, support vector machnes (SVM) []. However, n real applcatons, many problems are multclass problems. Therefore, generalzng SVM to deal wth multclass problems s stll one of mportant research actvtes n machne learnng. Currently, the usual practce s to convert a multclass problem nto a number of twoclass problems and then combne them n some way to realze classfcaton nto multple classes. Error Correctng Output Codes (ECOC) s one of the commonly used combnaton way [], whch s called ECOC SVMs. However, there s not a general codng method whch can generate approprate ECOC for any class number. Furthermore, the exstng codng strategy s based on the research on mathematcal features of code matrx, whch gnores the fundamentals of classfcaton, mang t dffcult to progress for ECOC SVMs and ther appled research. In ths study, t s beleved that, n real classfcaton problems, dfferent codng sequences of ECOC SVMs have dfferent meanngs. The performance of codng does not depend on ISSN: IJGDC Copyrght c 04 SERSC
2 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) code matrx tself; nstead, t depends on the performance of the real classfcaton problems correspondng to the code columns. Accordng to ths vewpont, we attempt to nvestgate ECOC SVMs from real classfcaton problems n ths study, whch ponts out the drecton for mprovng ECOC SVMs.. Code Matrx of ECOC and ts Correspondng Classfers. Code Matrx of ECOC ECOC s a codng matrx consstng of {0,} shown n Table, denoted as MQ S. In multclass problems, row Q represents the class number of samples, whle column S represents the number of classfers to be traned. When Mqs=(Mqs=0), ths sample s postve (negatve) for the qth class and the sth classfer fs. The worng process of ECOC s dvded nto two phases: tranng and classfcaton. In the tranng phase, the classfer f(x)=(f(x),,fs(x)) s traned accordng to the abovementoned prncple; whle n the classfcaton phase, for a new sample X, the dstances between output vectors and the class vectors are calculated. Then class wth the mnmum dstance s the classfcaton result, whch s gven by: K arg m n ( d ( M, f ( X )) () q.. Q q where K s the class of X, and d s the dstance functon. The Hammng Dstance (HD) s usually used: d ( M, f ( x )) q s m s g n ( f ) S q s s () Table. All Possble ECOC Columns for a 4Class Problem Class f f Code Word f 3 f 4 f 5 f 6 f 7 C C C C For ECOC, when the coded rows are the same, the classes correspondng to the rows cannot be dentfed; when the coded columns are the same, they correspond to the same classfer, therefore deletng a column does not affect the output; when the code of two columns are complementary, the outputs of ther correspondng classfers are complementary, therefore they are dentcal; columns of all 0 or all are mae no sense, because they cannot be used to tran classfers. In one word, an avalable ECOC must satsfy the followng condtons: ()The rows of the codng matrx are not correlated, and nether correlated nor complementary are the columns of the codng matrx; ()None of the columns s all 0 or all. (3)For a class problem, the codng length L must satsfy lo g L ; 68 Copyrght c 04 SERSC
3 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) Accordng to the codng theory, for an error correcton code wth mnmum HD d, [(d  )/ ] bts of error can be corrected. Therefore, for an output code wth error correcton ablty, the HD between code words should be larger than 3. Detterch proposed four commonly used ECOC codng methods [], ncludng Exhaustve Codes, Column Selecton from Exhaustve Codes, Randomzed Hll Clmbng Codes and BCH Codes. In addton, Crammer and Snger proposed the concept of contnuous codng [3]. Utschc proposed expectaton maxmzaton codng algorthm [4], n whch the ECOC s selected by constructng maxmzed objectve functon. Ludmla and Kuncheva used hybrdzaton and mutaton n evoluton algorthm to derve new ECOC codes from random ones [5]. The recent research about ECOC s a general codng method  searchng codng method whch was proposed n reference [6]. The method s not only sutable for problems of any class number, but also can automatcally generate alternatve codes accordng to dfferent crtera, ncludng class numbers and mnmum HDs. However, t cannot deal wth the problem caused by dentcal columns. For the evaluaton of codng performance, Francesco beleves that the performance of ECOC s related to many factors, ncludng: the smlarty of codng words, the performance of the classfers, the complexty of the real problems, the choce of classfers, and the correlaton of the codng columns, etc., [7]; Xa beleves the performance of ECOC s related to codng length, the mnmum HD between code words, and the dstrbuton order of the code words [8]. It can be seen from the above that the evaluaton of ECOC codng performance and the applcaton n classfcaton start from codng tself, whle attenton s seldom pad on classfcaton. Next, we ntroduce the ECOC SVMs n real classfcaton problems... SVM SVM s a machne learnng method based on statstcal learnng theory. To resolve the n pattern recognton problem, a calculable recognton functon y f(x ), x R, y , s found. For the gven samples (x, y ),(x, y ), (x, y ), x n R, y {, }, a hyperplane (decson surface) n needs to be found, namely, W x b 0,W R, b R, and the correspondng recognton functon s: f ( x ) sg n (( W x ) b ) (3) The decson surface should meet the followng constrants: y [ W x b ],,,, (4) The optmal decson surface should meet the requrement that the smallest dstance from the two classes of samples to the decson surface s the bggest, hence, the classfcaton problem becomes that the condton of Formula (4), namely: 0 should be met and the mnmum problem of m n : ( W ) W C (5) The frst tem n the formula maes the smallest dstance from the two classes of samples to the decson surface the bggest, whle the second tem maes the error the mnmum, and the constant C splts the dfference of two above. Ths optmzaton problem wth constrants Copyrght c 04 SERSC 69
4 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) can be resolved wth Lagrangan approach, and the correspondng classfcaton functon can be changed to: f ( x ) sg n ( y ( x x ) b ) (6) For the nonlnear separable condton, a nonlnear functon can be found, and then the data are mapped to a hgh dmensonal feature space, n whch an optmal hyperplane s establshed, and the correspondng classfcaton functon s as follows: f ( x ) sg n ( y ( ( x ) ( x )) b ) Only pont multplcaton algorthm K ( x, y ) ( x ) ( y ) n the hgh dmensonal feature space s consdered n SVM theory, n whch K ( x, y ) s called ernel functon, and the functon s not used drectly, hence, formula (7) can be transferred nto formula (8): (7) f ( x ) sg n ( y K ( x, x ) b ) The common ernel functon ncludes: lnear ernel functon, K ( x, y ) ( x y ) ; RBF ernel functon, ( x, y ) e x p ( x y / ). (8).3. ECOC SVMs Combnng SVMs wth ECOC to classfy multple classes, we have ECOC SVMs. The upper bound of the generalzaton error for ECOC SVMs s derved n reference [8] based on the concept of FatShatterng dmenson and coverng number. Assumng m samples can be correctly classfed by class ECOC SVMs, wth ECOC codng length beng L, mnmum HD between code words beng d, and the sorted SVM classfcaton ntervals n descendent order denoted by,,,, the generalzaton error ECOC SVMs, wth probablty at least L δ, s no larger than: M 3 0 R ( m ) M N K! ' D lo g ( 4 e m ) lo g ( 6 m ) lo g m (9) where D ' L, R s the mnmum radus of enclosure ball, M L ( d ) /, N s the number of codes wth codng length L and HD d between codes. Each group has K code words. It s beleved n [8] that: () Gven a fxed code length, the longer mnmum HD between codes, the better generalzaton ablty of ECOC SVMs; () Gven a fxed mnmum HD between codes, the longer code length, the worse generalzaton ablty of ECOC SVMs; (3) Once the code length and the mnmum HD between codes are fxed, there exsts optmal allocaton order for code words whch guarantees the ECOC SVMs the best generalzaton ablty. 70 Copyrght c 04 SERSC
5 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) The relaton between the generalzaton ablty of ECOC SVMs and the mnmum HD between codes and the code length s dscussed n [8]. However, no dscusson on the relaton between the mnmum HD between codes and the code length s gven. Furthermore, t s beleved n [8] that there do exst optmal codng sequence but no soluton s provded accordng to codng tself. Clearly, t s dffcult to fnd a way to determne the code length and the code sequences from a mathematc pont of vew. Exhaustve search s defntely unfeasble and t does not provde a reasonable explanaton to codng. Intutvely, the answer should be found n the classfcaton problems themselves. 3. New Understandngs about ECOC SVMs Analyzng formula (9) agan, one nows that D, M and N affect the upper bound of the generalzaton error of ECOC SVMs. In formula (9), M L d, the mnmum HD d s related to code length L. Generally, the larger L, the larger d. However the ncrement of d s less than or equal to that of L. Therefore, M s nondecreasng; N s also related to L. It ncreases when L ncreases. D ' L, wth the ncrease of code length L, D s ncreasng. Therefore, the generalzaton ablty of ECOC SVMs decreases when L ncreases. In ths study, we beleve that ECOC SVMs should have enough SubSVMs for decson. That s to say, L should be bg enough. However, a bgger L may harm the performance of ECOC SVMs. Thus, the value of L should be a compromse. When L s determned, the effect of D on the generalzaton ablty of ECOC SVMs s major. Select L SubSVMs wth good generalzaton ablty, then the generalzaton ablty should be good f one constructs ECOC SVMs wth these SubSVMs. If t s mpossble to evaluate the generalzaton ablty of each SubSVM, or the dfference between each s nsgnfcant, the effect of M and N on the performance of ECOC SVMs, whch s the concluson of reference [8]. A votng process s used to vvdly descrbe the above analyss. The essence of ECOC SVMs s to tran a number of twoclass SVMs, then determne the class of an unnown sample accordng to the classfcaton results of these twoclass SVMs. Usng mnmum HD to determne the class of a sample s equvalent to votng. In the votng stage, each SubSVM votes for a number of classes whch t supports; n the callng stage, the sample class s the class whch most corresponds to the results of SubSVMs. Each column of the codes corresponds to a SubSVM, therefore, the process of constructng ECOC determnes whch of the SubSVMs have the votng rght. Clearly, t s mportant to gve votng rght to those Sub SVMs wth good generalzaton ablty. Thus, the codng problem s actually how to construct an optmal votng machne consstng of a number of twoclass SVMs, where how to select SubSVMs wth good generalzaton ablty and how to determne the number of SubSVMs tang part n votng are two mportant factors. Next, experments are used to valdate the vewpont. It s ponted out n [] that VC dmenson and the classfcaton nterval when lnearly separable are the crtera for the generalzaton ablty of an SVM. However, t s dffcult to determne the VC dmenson. Therefore VC dmenson s dffcult to deal wth and apply. The classfcaton nterval s a relable crteron for evaluatng the generalzaton ablty of an SVM, but t needs the precondton that the SVM can lnearly separate samples, whch s usually dffcult to satsfy. When samples are not lnearly separable but are lnearly separable after beng mapped nto a hgh dmensonal space, the generalzaton ablty of an SVM s descrbed by the classfcaton nterval n hgh dmensonal space. In ths stuaton, the classfcaton nterval of an SVM s / W, where W s the normal vector of the Copyrght c 04 SERSC 7
6 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) classfcaton hyperplane of the SVM, whch s calculable. However t s mpossble to drectly calculate W. Notcng the dualty of (6) and (8), one can obtan: (0) j j j W C y y K ( x, x ), j When samples are completely separable, C 0, then (0) s smplfed as: () j j j W y y K ( x, x ), j W can be calculated accordng to (), thereby the classfcaton nterval n hgh dmensonal space can also be obtaned. When samples are partally separable, we use the error rate E of the cross valdaton to evaluate the generalzaton ablty of SVMs. The smaller E, the better generalzaton ablty. To reduce the msclassfcaton, the classfcaton nterval s also consdered when usng E as a evaluaton crteron. But the classfcaton nterval now ncludes the msclassfed samples. The effect of msclassfed samples should be elmnated when calculatng. Data sets from UCI database are selected for test, ncludng Segment, Landsat, Optdgts, Zoo, Page Blocs, etc., Lnear ernel and RBF ernel are used n the test. The process of the test s as follow: Gven classes of samples, all the ECOC code columns are constructed by exhaustve method, totalng columns. For each column of code, tran the SubSVMs. Then sort the code columns by the generalzaton ablty of SubSVMs, accordng to the followng rules: () When samples are lnearly separable, the lnear ernels are chosen. They have lower VC dmensons and better generalzaton ablty compared wth RBF ernels; () When samples are lnearly nseparable but separable f RBF ernels are used, the RBF ernels are chosen; (3) When samples are completely nseparable, ernel functons wth hgher accuraces are chosen; (4) When samples are separable, sort the classfcaton nterval by descendng order; (5) When samples are nseparable, sort the error rate E by ascendng order, meantme are referred. A number of code columns n postve sequence are chosen from the sorted ECOC code columns to tran the ECOC SVMs, then the same number of code columns n reverse order are chosen as comparson group. The orgnal order of the exhaustve code s ept unchanged. The same number of code columns are successvely chosen as reference groups. The expermental results of the selected data sets are bascally the same. Tang the Segment data set as an example, the results are shown n Fgure. There are 7 classes n Segment data set, wth each sample havng 9 features. Each class provdes 30 tranng samples and 300 test samples. The code length ranges from 3 to 63. In the experments, ECOC SVMs consstng of 0 to 63 code columns are tested. RBF ernels have hgher accuracy, so they are chosen n the experments. To facltate the test, same parameters are used for all SubSVMs. 7 Copyrght c 04 SERSC
7 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) Predcton accuracy 0.95 the relatonshp between ECOC codng sequence and the correspondng predcton accuracy Code Length forward sequence reverse sequence orgnal sequence Fgure. Relatonshp between ECOC SVMs Encodng Method and ther Predcton Accuracy for Segment Data Set It can be seen n Fgure that the predcton accuracy of ECOC forward sequences s much hgher than that of reverse sequences, whle the orgnal sequences have the medum accuracy. The orgnal sequences can be vewed as the predcton of random ECOC, whle the reverse sequences have the worst predcton results. The forward sequences have the best predcton results. In addton, the more overlappng code between forward and reverse sequences, the closer predcton accuraces they have. When code length ncreases to a certan degree, the predcton accuracy of forward sequences decreases and becomes stable, whle the accuracy of orgnal sequences ncreases wth fluctuatons and fnally becomes stable. However the accuracy of reverse sequences eeps ncreasng. The results suggest that when code length ncreases, f the generalzaton ablty of the SVMs correspondng to the newly added columns are strong, the codng performance mproves, le reverse sequences; Conversely, f the generalzaton ablty of the SVMs s wea, the codng performance degrades, le forward sequences; n orgnal sequences, code lengths are short at the begnnng, whch maes the codng performance bad, however wth the ncrease of code lengths, the mnmum HD between codes ncreases, mprovng the generalzaton ablty. But f the code lengths stll ncrease, more code columns wth wea generalzaton ablty exst, whch stops the overall performance from ncreasng. It s beleved n [8] that the generalzaton ablty of ECOC SVMs depends on the frst [L (d  )/ ] hgh performance SVMs, rrelatve to the rest (d)/ SVMs. However we further show that when generalzaton ablty s good, nsuffcent codng numbers wll also affect the generalzaton ablty of ECOC SVMs. In ths stuaton, the generalzaton ablty of ECOC SVMs s relatve to the SubSVMs wth bad generalzaton ablty. More mportantly, a codng strategy s derved n ths study showng how to construct ECOC wth good generalzaton ablty. It can be seen n Fgure that, the performance of ECOC SVMs nether necessarly ncreases wth the ncrease of the mnmum HD, nor wth the ncrease of code length. Instead, t has a complex relatonshp wth both of them. Copyrght c 04 SERSC 73
8 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) Fgure. Relatonshp between Predcton Accuracy, Code Length, and the Mnmum Hammng Dstance The vewpont n ths study s valdated through real tests. That s, the generalzaton ablty of SVMs correspondng to code columns has the most sgnfcant effect on ECOC performance, whle code lengths and mnmum HD between codes are both mathematcal features represented by codng. The ey factor n mprovng the performance of ECOC SVMs s to fnd SubSVMs wth good generalzaton ablty. When mpossble to fnd such Sub SVMs or the generalzaton abltes of all SubSVMs are the same, the performance of ECOC SVMs can be consdered from the codng pont of vew. Then code length, mnmum HD between codes, allocaton order of codes, correlatons between code columns can be the evaluaton crtera of the generalzaton ablty of ECOC SVMs, whch s dscussed n [8]. The exhaustve codng method can assure the code columns are nether correlated nor complementary. However when code length s short, same columns may exst n ECOC, mae t mpossble to determne the class of a sample. In ths stuaton, remedal measures should be taen, that s, tranng addtonal SVM classfers correspondng to the dentcal code words, n order to judge the decson results of ECOC SVMs. In Fgure, there are cases n whch the mnmum HD between codes s 0 when the reverse sequence and the orgnal sequence are both short, mplyng that there are dentcal codes. Strctly speang, the ECOC s wrong n these cases. However for convenence, ths part of code s reserved after tang measures to deal wth t, mared by n Fgure. 4. Conclusons and Dscussons Startng from the essence of problems, ECOC SVMs s analyzed n ths study. New constructng method s proposed. The man concluson and problem are as below:. The performance of ECOC SVMs depends on the performance of ts correspondng Sub SVMs, whle the mathematcal features represented by codng are secondary. When mpossble to evaluate the performances of SubSVMs or the performances are the same, the code lengths, the mnmum HD between codes, the allocaton order of code words, and the 74 Copyrght c 04 SERSC
9 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) correlatons between codes can be the crtera of ECOC. For the error correcton ablty, t s beleved n ths study that for optmal classfers, less error s prmary, whle error correcton s secondary.. What needs to further solve s that: () currently, there are stll no exact theory about the evaluaton of the SVM performance because ernel functons, parameters and sample spaces are dfferent. Also, there s stll no exact theory about the comparablty of the generalzaton ablty of each SubSVM; () What s the approprate code length of ECOC SVMs? What s the relaton between the code length of ECOC SVMs and the generalzaton ablty of each SubSVM? And how to convenently and rapdly construct a reasonable code matrx? These problems are stll to be nvestgated. From the ntal results, code can be short for SubSVMs wth better generalzaton abltes, or qualty s more mportant than quantty; when the performance of SubSVMs s mpossble to evaluate, code length should be longer, or compensate qualty by ncreasng quantty. But the code length should be approprate, by no means the longer the better. Acnowledgements Ths wor was supported by a grant from Natural Scentfc Fund of Chna (No ) and a Project Funded by the Prorty Academc Program Development of Jangsu Hgher Educaton Insttutons. References [] V. N. Vapn, The Nature of Statstcal Learnng Theory, Sprnger, New Yor, USA, (995). [] T. Detterch and G. Bar, Solvng multclass learnng problems va errorcorrectng output codes, Journal of Artfcal Intellgence Research, vol., (995), pp [3] K. Crammer and Y. Snger, On the learnablty and desgn of output codes for multclass problems, Proc. of the 3th Annual Conf. on Computatonal Learnng Theory, (000), pp [4] W. Utschc and W. Wechselberger, Stochastc organzaton of output codes n multclass learnng problems, Neural Computng, vol. 3, no. 5, (00), pp [5] K. Ludmla I, Usng dversty measures for generatng errorcorrectng output codes n classfer ensembles, Pattern Recognton Letters, vol. 6, no., (005), pp [6] Y. Jang, Q. Zhao and X. Yang, A Search Codng Method and Its Applcaton n Supervsed Classfcaton, Journal of Software, (In Chnese), vol. 6, no. 06, (005), pp [7] F. Masull and G. Valentn, An expermental analyss of the dependence among codeword bt errors n ECOC learnng machnes, Neuro computng, vol. 57, (004), pp [8] X. Jantao and H. Mngy, Multclass Classfcaton Usng Support Vector Machnes (SVMs) Combned wth ErrorCorrectng Codes (ECCs), Journal of Northwestern Polytechncal Unversty, (In Chnese), vol., no. 4, (003), pp Authors Zhgang Yan receved B.Sc. degree from Chna Unversty of Mnng and Technology n 997 and Ph.D. degree from Chna Unversty of Mnng and Technology n 007. He s currently a assocate professor at faculty of Chna Unversty of Mnng and Technology, Chna. Hs feld of nterest s spatotemporal data mnng and nowledge dscoverng. Yuanxuan Yang, Master student, receved B.Sc. degree n Geographc Informaton System n 03 from Chna Unversty of Mnng and Technology. Now he study n Chna Unversty of Mnng and Technology, supervsed by Zhgang Yan. Copyrght c 04 SERSC 75
10 Internatonal Journal of Grd and Dstrbuted Computng Vol.7, No. (04) 76 Copyrght c 04 SERSC
The Development of Web Log Mining Based on ImproveKMeans Clustering Analysis
The Development of Web Log Mnng Based on ImproveKMeans Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.
More informationAn Alternative Way to Measure Private Equity Performance
An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate
More informationForecasting 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 informationFeature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College
Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure
More informationSupport 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 informationStudy on CET4 Marks in China s Graded English Teaching
Study on CET4 Marks n Chna s Graded Englsh Teachng CHE We College of Foregn Studes, Shandong Insttute of Busness and Technology, P.R.Chna, 264005 Abstract: Ths paper deploys Logt model, and decomposes
More informationCommunication Networks II Contents
8 / 1  Communcaton Networs II (Görg)  www.comnets.unbremen.de Communcaton Networs II Contents 1 Fundamentals of probablty theory 2 Traffc n communcaton networs 3 Stochastc & Marovan Processes (SP
More informationDesign of Output Codes for Fast Covering Learning using Basic Decomposition Techniques
Journal of Computer Scence (7): 56557, 6 ISSN 5966 6 Scence Publcatons Desgn of Output Codes for Fast Coverng Learnng usng Basc Decomposton Technques Aruna Twar and Narendra S. Chaudhar, Faculty of Computer
More informationWhat 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 informationNonlinear data mapping by neural networks
Nonlnear data mappng by neural networks R.P.W. Dun Delft Unversty of Technology, Netherlands Abstract A revew s gven of the use of neural networks for nonlnear mappng of hgh dmensonal data on lower dmensonal
More informationForecasting 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 informationLogistic 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 informationGender Classification for RealTime Audience Analysis System
Gender Classfcaton for RealTme Audence Analyss System Vladmr Khryashchev, Lev Shmaglt, Andrey Shemyakov, Anton Lebedev Yaroslavl State Unversty Yaroslavl, Russa vhr@yandex.ru, shmaglt_lev@yahoo.com, andrey.shemakov@gmal.com,
More informationL10: 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 informationInstitute 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 informationStudy on Model of Risks Assessment of Standard Operation in Rural Power Network
Study on Model of Rsks Assessment of Standard Operaton n Rural Power Network Qngj L 1, Tao Yang 2 1 Qngj L, College of Informaton and Electrcal Engneerng, Shenyang Agrculture Unversty, Shenyang 110866,
More informationImproved SVM in Cloud Computing Information Mining
Internatonal Journal of Grd Dstrbuton Computng Vol.8, No.1 (015), pp.3340 http://dx.do.org/10.1457/jgdc.015.8.1.04 Improved n Cloud Computng Informaton Mnng Lvshuhong (ZhengDe polytechnc college JangSu
More informationPerformance Management and Evaluation Research to University Students
631 A publcaton of CHEMICAL ENGINEERING TRANSACTIONS VOL. 46, 2015 Guest Edtors: Peyu Ren, Yancang L, Hupng Song Copyrght 2015, AIDIC Servz S.r.l., ISBN 9788895608372; ISSN 22839216 The Italan Assocaton
More informationCan Auto Liability Insurance Purchases Signal Risk Attitude?
Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? ChuShu L Department of Internatonal Busness, Asa Unversty, Tawan ShengChang
More informationANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING
ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING Matthew J. Lberatore, Department of Management and Operatons, Vllanova Unversty, Vllanova, PA 19085, 6105194390,
More informationImplementation and Evaluation of a Random Forest Machine Learning Algorithm
Implementaton and Evaluaton of a Random Forest Machne Learnng Algorthm Vachaslau Sazonau Unversty of Manchester, Oxford Road, Manchester, M13 9PL,UK sazonauv@cs.manchester.ac.uk Abstract hs work s amed
More informationSingle 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 informationModule 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur
Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..
More informationLETTER IMAGE RECOGNITION
LETTER IMAGE RECOGNITION 1. Introducton. 1. Introducton. Objectve: desgn classfers for letter mage recognton. consder accuracy and tme n takng the decson. 20,000 samples: Startng set: mages based on 20
More informationThe Greedy Method. Introduction. 0/1 Knapsack Problem
The Greedy Method Introducton We have completed data structures. We now are gong to look at algorthm desgn methods. Often we are lookng at optmzaton problems whose performance s exponental. For an optmzaton
More informationA 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):18841889 Research Artcle ISSN : 09757384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel
More informationAn InterestOriented Network Evolution Mechanism for Online Communities
An InterestOrented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne
More informationLinear Regression, Regularization BiasVariance Tradeoff
HTF: Ch3, 7 B: Ch3 Lnear Regresson, Regularzaton BasVarance Tradeoff Thanks to C Guestrn, T Detterch, R Parr, N Ray 1 Outlne Lnear Regresson MLE = Least Squares! Bass functons Evaluatng Predctors Tranng
More information320 The Internatonal Arab Journal of Informaton Technology, Vol. 5, No. 3, July 2008 Comparsons Between Data Clusterng Algorthms Osama Abu Abbas Computer Scence Department, Yarmouk Unversty, Jordan Abstract:
More informationA Performance Analysis of View Maintenance Techniques for Data Warehouses
A Performance Analyss of Vew Mantenance Technques for Data Warehouses Xng Wang Dell Computer Corporaton Round Roc, Texas Le Gruenwald The nversty of Olahoma School of Computer Scence orman, OK 739 Guangtao
More informationIMPACT ANALYSIS OF A CELLULAR PHONE
4 th ASA & μeta Internatonal Conference IMPACT AALYSIS OF A CELLULAR PHOE We Lu, 2 Hongy L Bejng FEAonlne Engneerng Co.,Ltd. Bejng, Chna ABSTRACT Drop test smulaton plays an mportant role n nvestgatng
More informationTHE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION
Internatonal Journal of Electronc Busness Management, Vol. 3, No. 4, pp. 3030 (2005) 30 THE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION YuMn Chang *, YuCheh
More informationQuestions that we may have about the variables
Antono Olmos, 01 Multple Regresson Problem: we want to determne the effect of Desre for control, Famly support, Number of frends, and Score on the BDI test on Perceved Support of Latno women. Dependent
More informationFace 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 informationProject Networks With MixedTime Constraints
Project Networs Wth MxedTme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa
More informationOn Mean Squared Error of Hierarchical Estimator
S C H E D A E I N F O R M A T I C A E VOLUME 0 0 On Mean Squared Error of Herarchcal Estmator Stans law Brodowsk Faculty of Physcs, Astronomy, and Appled Computer Scence, Jagellonan Unversty, Reymonta
More information"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *
Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC
More informationUsing Mixture Covariance Matrices to Improve Face and Facial Expression Recognitions
Usng Mxture Covarance Matrces to Improve Face and Facal Expresson Recogntons Carlos E. homaz, Duncan F. Glles and Raul Q. Fetosa 2 Imperal College of Scence echnology and Medcne, Department of Computng,
More informationSearching for Interacting Features for Spam Filtering
Searchng for Interactng Features for Spam Flterng Chuanlang Chen 1, YunChao Gong 2, Rongfang Be 1,, and X. Z. Gao 3 1 Department of Computer Scence, Bejng Normal Unversty, Bejng 100875, Chna 2 Software
More informationOptimization Model of Reliable Data Storage in Cloud Environment Using Genetic Algorithm
Internatonal Journal of Grd Dstrbuton Computng, pp.175190 http://dx.do.org/10.14257/gdc.2014.7.6.14 Optmzaton odel of Relable Data Storage n Cloud Envronment Usng Genetc Algorthm Feng Lu 1,2,3, Hatao
More informationRESEARCH ON DUALSHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo.
ICSV4 Carns Australa 9 July, 007 RESEARCH ON DUALSHAKER SINE VIBRATION CONTROL Yaoq FENG, Hanpng QIU Dynamc Test Laboratory, BISEE Chna Academy of Space Technology (CAST) yaoq.feng@yahoo.com Abstract
More informationLecture 18: Clustering & classification
O CPS260/BGT204. Algorthms n Computatonal Bology October 30, 2003 Lecturer: Pana K. Agarwal Lecture 8: Clusterng & classfcaton Scrbe: Daun Hou Open Problem In HomeWor 2, problem 5 has an open problem whch
More informationA Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy Scurve Regression
Novel Methodology of Workng Captal Management for Large Publc Constructons by Usng Fuzzy Scurve Regresson ChengWu Chen, Morrs H. L. Wang and TngYa Hseh Department of Cvl Engneerng, Natonal Central Unversty,
More informationSensitivity Analysis in a Generic MultiAttribute Decision Support System
Senstvty Analyss n a Generc MultAttrbute Decson Support System Sxto RíosInsua, Antono Jménez and Alfonso Mateos Department of Artfcal Intellgence, Madrd Techncal Unversty Campus de Montegancedo s/n,
More informationIdentifying Workloads in Mixed Applications
, pp.395400 http://dx.do.org/0.4257/astl.203.29.8 Identfyng Workloads n Mxed Applcatons Jeong Seok Oh, Hyo Jung Bang, Yong Do Cho, Insttute of Gas Safety R&D, Korea Gas Safety Corporaton, ShghungSh,
More information1 Example 1: Axisaligned 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 informationConversion between the vector and raster data structures using Fuzzy Geographical Entities
Converson between the vector and raster data structures usng Fuzzy Geographcal Enttes Cdála Fonte Department of Mathematcs Faculty of Scences and Technology Unversty of Combra, Apartado 38, 3 454 Combra,
More informationBayesian Network Based Causal Relationship Identification and Funding Success Prediction in P2P Lending
Proceedngs of 2012 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 25 (2012) (2012) IACSIT Press, Sngapore Bayesan Network Based Causal Relatonshp Identfcaton and Fundng Success
More informationA New Task Scheduling Algorithm Based on Improved Genetic Algorithm
A New Task Schedulng Algorthm Based on Improved Genetc Algorthm n Cloud Computng Envronment Congcong Xong, Long Feng, Lxan Chen A New Task Schedulng Algorthm Based on Improved Genetc Algorthm n Cloud Computng
More informationData Broadcast on a MultiSystem Heterogeneous Overlayed Wireless Network *
JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 819840 (2008) Data Broadcast on a MultSystem Heterogeneous Overlayed Wreless Network * Department of Computer Scence Natonal Chao Tung Unversty Hsnchu,
More informationResearch Article Enhanced TwoStep Method via Relaxed Order of αsatisfactory Degrees for Fuzzy Multiobjective Optimization
Hndaw Publshng Corporaton Mathematcal Problems n Engneerng Artcle ID 867836 pages http://dxdoorg/055/204/867836 Research Artcle Enhanced TwoStep Method va Relaxed Order of αsatsfactory Degrees for Fuzzy
More informationCS 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 informationCHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol
CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL
More informationRobust Design of Public Storage Warehouses. Yeming (Yale) Gong EMLYON Business School
Robust Desgn of Publc Storage Warehouses Yemng (Yale) Gong EMLYON Busness School Rene de Koster Rotterdam school of management, Erasmus Unversty Abstract We apply robust optmzaton and revenue management
More informationDescriptive Models. Cluster Analysis. Example. General Applications of Clustering. Examples of Clustering Applications
CMSC828G Prncples of Data Mnng Lecture #9 Today s Readng: HMS, chapter 9 Today s Lecture: Descrptve Modelng Clusterng Algorthms Descrptve Models model presents the man features of the data, a global summary
More informationA Computer Technique for Solving LP Problems with Bounded Variables
Dhaka Unv. J. Sc. 60(2): 163168, 2012 (July) A Computer Technque for Solvng LP Problems wth Bounded Varables S. M. Atqur Rahman Chowdhury * and Sanwar Uddn Ahmad Department of Mathematcs; Unversty of
More informationMining Feature Importance: Applying Evolutionary Algorithms within a Webbased Educational System
Mnng Feature Importance: Applyng Evolutonary Algorthms wthn a Webbased Educatonal System Behrouz MINAEIBIDGOLI 1, and Gerd KORTEMEYER 2, and Wllam F. PUNCH 1 1 Genetc Algorthms Research and Applcatons
More informationAn Analysis of Factors Influencing the SelfRated Health of Elderly Chinese People
Open Journal of Socal Scences, 205, 3, 520 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/ss http://dx.do.org/0.4236/ss.205.35003 An Analyss of Factors Influencng the SelfRated Health of
More informationSorting Online Reviews by Usefulness Based on the VIKOR Method
Assocaton or Inormaton Systems AIS Electronc Lbrary (AISeL) Eleventh Wuhan Internatonal Conerence on e Busness Wuhan Internatonal Conerence on ebusness 5262012 Sortng Onlne Revews by Useulness Based
More information8.5 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 informationNEUROFUZZY INFERENCE SYSTEM FOR ECOMMERCE WEBSITE EVALUATION
NEUROFUZZY INFERENE SYSTEM FOR EOMMERE WEBSITE EVALUATION Huan Lu, School of Software, Harbn Unversty of Scence and Technology, Harbn, hna Faculty of Appled Mathematcs and omputer Scence, Belarusan State
More informationUsing Supervised Clustering Technique to Classify Received Messages in 137 Call Center of Tehran City Council
Usng Supervsed Clusterng Technque to Classfy Receved Messages n 137 Call Center of Tehran Cty Councl Mahdyeh Haghr 1*, Hamd Hassanpour 2 (1) Informaton Technology engneerng/ecommerce, Shraz Unversty (2)
More informationGrid Resource Selection Optimization with Guarantee Quality of Service by GAPSO
Australan Journal of Basc and Appled Scences, 5(11): 21392145, 2011 ISSN 19918178 Grd Resource Selecton Optmzaton wth Guarantee Qualty of Servce by GAPSO 1 Hossen Shrgah and 2 Nameh Danesh 1 Islamc Azad
More informationData Mining from the Information Systems: Performance Indicators at Masaryk University in Brno
Data Mnng from the Informaton Systems: Performance Indcators at Masaryk Unversty n Brno Mkuláš Bek EUA Workshop Strasbourg, 12 December 2006 1 Locaton of Brno Brno EUA Workshop Strasbourg, 12 December
More informationSupport vector domain description
Pattern Recognton Letters 20 (1999) 1191±1199 www.elsever.nl/locate/patrec Support vector doman descrpton Davd M.J. Tax *,1, Robert P.W. Dun Pattern Recognton Group, Faculty of Appled Scence, Delft Unversty
More informationNumber of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000
Problem Set 5 Solutons 1 MIT s consderng buldng a new car park near Kendall Square. o unversty funds are avalable (overhead rates are under pressure and the new faclty would have to pay for tself from
More informationA DATA MINING APPLICATION IN A STUDENT DATABASE
JOURNAL OF AERONAUTICS AND SPACE TECHNOLOGIES JULY 005 VOLUME NUMBER (5357) A DATA MINING APPLICATION IN A STUDENT DATABASE Şenol Zafer ERDOĞAN Maltepe Ünversty Faculty of Engneerng BüyükbakkalköyIstanbul
More informationAN APPOINTMENT ORDER OUTPATIENT SCHEDULING SYSTEM THAT IMPROVES OUTPATIENT EXPERIENCE
AN APPOINTMENT ORDER OUTPATIENT SCHEDULING SYSTEM THAT IMPROVES OUTPATIENT EXPERIENCE YuL Huang Industral Engneerng Department New Mexco State Unversty Las Cruces, New Mexco 88003, U.S.A. Abstract Patent
More informationGraph Theory and Cayley s Formula
Graph Theory and Cayley s Formula Chad Casarotto August 10, 2006 Contents 1 Introducton 1 2 Bascs and Defntons 1 Cayley s Formula 4 4 Prüfer Encodng A Forest of Trees 7 1 Introducton In ths paper, I wll
More informationECE544NA Final Project: Robust Machine Learning Hardware via Classifier Ensemble
1 ECE544NA Fnal Project: Robust Machne Learnng Hardware va Classfer Ensemble Sa Zhang, szhang12@llnos.edu Dept. of Electr. & Comput. Eng., Unv. of Illnos at UrbanaChampagn, Urbana, IL, USA Abstract In
More informationThe 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 informationSIX WAYS TO SOLVE A SIMPLE PROBLEM: FITTING A STRAIGHT LINE TO MEASUREMENT DATA
SIX WAYS TO SOLVE A SIMPLE PROBLEM: FITTING A STRAIGHT LINE TO MEASUREMENT DATA E. LAGENDIJK Department of Appled Physcs, Delft Unversty of Technology Lorentzweg 1, 68 CJ, The Netherlands Emal: e.lagendjk@tnw.tudelft.nl
More informationA ChiSquareTest for Word Importance Differentiation in Text Classification
011 Internatonal Conference on Informaton and Electroncs Engneerng IPCSIT vol.6 (011) (011) IACSIT Press, Sngapore A ChSquareTest for Word Importance Dfferentaton n Text Classfcaton Phayung Meesad 1,
More informationA ReplicationBased and Fault Tolerant Allocation Algorithm for Cloud Computing
A ReplcatonBased and Fault Tolerant Allocaton Algorthm for Cloud Computng Tork Altameem Dept of Computer Scence, RCC, Kng Saud Unversty, PO Box: 28095 11437 RyadhSaud Araba Abstract The very large nfrastructure
More informationLecture 2: Single Layer Perceptrons Kevin Swingler
Lecture 2: Sngle Layer Perceptrons Kevn Sngler kms@cs.str.ac.uk Recap: McCullochPtts Neuron Ths vastly smplfed model of real neurons s also knon as a Threshold Logc Unt: W 2 A Y 3 n W n. A set of synapses
More informationRiskbased Fatigue Estimate of Deep Water Risers  Course Project for EM388F: Fracture Mechanics, Spring 2008
Rskbased 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 informationLogistic Regression. Steve Kroon
Logstc Regresson Steve Kroon Course notes sectons: 24.324.4 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 informationSVM Tutorial: Classification, Regression, and Ranking
SVM Tutoral: Classfcaton, Regresson, and Rankng Hwanjo Yu and Sungchul Km 1 Introducton Support Vector Machnes(SVMs) have been extensvely researched n the data mnng and machne learnng communtes for the
More informationOpen Access A Load Balancing Strategy with Bandwidth Constraint in Cloud Computing. Jing Deng 1,*, Ping Guo 2, Qi Li 3, Haizhu Chen 1
Send Orders for Reprnts to reprnts@benthamscence.ae The Open Cybernetcs & Systemcs Journal, 2014, 8, 115121 115 Open Access A Load Balancng Strategy wth Bandwdth Constrant n Cloud Computng Jng Deng 1,*,
More informationTime Domain simulation of PD Propagation in XLPE Cables Considering Frequency Dependent Parameters
Internatonal Journal of Smart Grd and Clean Energy Tme Doman smulaton of PD Propagaton n XLPE Cables Consderng Frequency Dependent Parameters We Zhang a, Jan He b, Ln Tan b, Xuejun Lv b, HongJe L a *
More informationHYPOTHESIS TESTING OF PARAMETERS FOR ORDINARY LINEAR CIRCULAR REGRESSION
HYPOTHESIS TESTING OF PARAMETERS FOR ORDINARY LINEAR CIRCULAR REGRESSION Abdul Ghapor Hussn Centre for Foundaton Studes n Scence Unversty of Malaya 563 KUALA LUMPUR Emal: ghapor@umedumy Abstract Ths paper
More informationFinancial Mathemetics
Fnancal Mathemetcs 15 Mathematcs Grade 12 Teacher Gude Fnancal Maths Seres Overvew In ths seres we am to show how Mathematcs can be used to support personal fnancal decsons. In ths seres we jon Tebogo,
More informationThe Application of Fractional Brownian Motion in Option Pricing
Vol. 0, No. (05), pp. 738 http://dx.do.org/0.457/jmue.05.0..6 The Applcaton of Fractonal Brownan Moton n Opton Prcng Qngxn Zhou School of Basc Scence,arbn Unversty of Commerce,arbn zhouqngxn98@6.com
More informationCalculating the high frequency transmission line parameters of power cables
< ' Calculatng the hgh frequency transmsson lne parameters of power cables Authors: Dr. John Dcknson, Laboratory Servces Manager, N 0 RW E B Communcatons Mr. Peter J. Ncholson, Project Assgnment Manager,
More information1 Approximation Algorithms
CME 305: Dscrete Mathematcs and Algorthms 1 Approxmaton Algorthms In lght of the apparent ntractablty of the problems we beleve not to le n P, t makes sense to pursue deas other than complete solutons
More informationCausal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting
Causal, Explanatory Forecastng Assumes causeandeffect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of
More informationTraditional versus Online Courses, Efforts, and Learning Performance
Tradtonal versus Onlne Courses, Efforts, and Learnng Performance KuangCheng Tseng, Department of Internatonal Trade, ChungYuan Chrstan Unversty, Tawan ShanYng Chu, Department of Internatonal Trade,
More informationImproved Mining of Software Complexity Data on Evolutionary Filtered Training Sets
Improved Mnng of Software Complexty Data on Evolutonary Fltered Tranng Sets VILI PODGORELEC Insttute of Informatcs, FERI Unversty of Marbor Smetanova ulca 17, SI2000 Marbor SLOVENIA vl.podgorelec@unmb.s
More informationDetecting Credit Card Fraud using Periodic Features
Detectng Credt Card Fraud usng Perodc Features Alejandro Correa Bahnsen, Djamla Aouada, Aleksandar Stojanovc and Björn Ottersten Interdscplnary Centre for Securty, Relablty and Trust Unversty of Luxembourg,
More informationDesign and Development of a Security Evaluation Platform Based on International Standards
Internatonal Journal of Informatcs Socety, VOL.5, NO.2 (203) 780 7 Desgn and Development of a Securty Evaluaton Platform Based on Internatonal Standards Yuj Takahash and Yoshm Teshgawara Graduate School
More informationRecurrence. 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 informationAnswer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy
4.02 Quz Solutons Fall 2004 MultpleChoce Questons (30/00 ponts) Please, crcle the correct answer for each of the followng 0 multplechoce questons. For each queston, only one of the answers s correct.
More informationFisher 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 informationAn artificial Neural Network approach to monitor and diagnose multiattribute quality control processes. S. T. A. Niaki*
Journal of Industral Engneerng Internatonal July 008, Vol. 4, No. 7, 04 Islamc Azad Unversty, South Tehran Branch An artfcal Neural Network approach to montor and dagnose multattrbute qualty control processes
More informationPlanning for Marketing Campaigns
Plannng for Marketng Campagns Qang Yang and Hong Cheng Department of Computer Scence Hong Kong Unversty of Scence and Technology Clearwater Bay, Kowloon, Hong Kong, Chna (qyang, csch)@cs.ust.hk Abstract
More informationSoftware project management with GAs
Informaton Scences 177 (27) 238 241 www.elsever.com/locate/ns Software project management wth GAs Enrque Alba *, J. Francsco Chcano Unversty of Málaga, Grupo GISUM, Departamento de Lenguajes y Cencas de
More informationHow Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence
1 st Internatonal Symposum on Imprecse Probabltes and Ther Applcatons, Ghent, Belgum, 29 June 2 July 1999 How Sets of Coherent Probabltes May Serve as Models for Degrees of Incoherence Mar J. Schervsh
More informationEnabling P2P Oneview Multiparty Video Conferencing
Enablng P2P Onevew Multparty Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract MultParty Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P
More informationGenetic algorithm for searching for critical slip surface in gravity dams based on stress fields CHEN Jianyun 1, WANG Shu 2, XU Qiang 3, LI Jing 4
Advanced Materals Research Onlne: 2030904 ISSN: 6628985, Vol. 790, pp 4649 do:0.4028/www.scentfc.net/amr.790.46 203 Trans Tech Publcatons, Swtzerland Genetc algorthm for searchng for crtcal slp surface
More informationPeriod and Deadline Selection for Schedulability in RealTime Systems
Perod and Deadlne Selecton for Schedulablty n RealTme Systems Thdapat Chantem, Xaofeng Wang, M.D. Lemmon, and X. Sharon Hu Department of Computer Scence and Engneerng, Department of Electrcal Engneerng
More information