Information Sciences

Size: px
Start display at page:

Download "Information Sciences"

Transcription

1 Ifrmati Scieces 181 (2011) Ctets lists available at ScieceDirect Ifrmati Scieces jural hmepage: Parameterized attribute reducti with aussia kerel based fuzzy rugh sets Degag Che a,, Qighua Hu b, Ygpig Yag a a Nrth Chia Electric Pwer Uiversity, Beijig , PR Chia b Harbi Istitute f Techlgy, Harbi , PR Chia article if abstract Article histry: Received 15 September 2008 Received i revised frm 24 Jue 2011 Accepted 5 July 2011 Available lie 23 July 2011 Keywrds: aussia kerels Fuzzy rugh sets Feature selecti Disceribility matrix Fuzzy rugh sets are csidered as a effective tl t deal with ucertaity i data aalysis, ad fuzzy similarity relatis are used i fuzzy rugh sets t calculate similarity betwee bjects. O the ther had i kerel tricks, a kerel maps data it a higher dimesial feature space where the resultig structure f the learig task is liearly separable, while the kerel is the ier prduct f this feature space ad ca als be viewed as a similarity fucti. It has bee reprted there is a verlap betwee family f kerels ad cllecti f fuzzy similarity relatis. This fact mtivates the idea i this paper t use sme kerels as fuzzy similarity relatis ad develp kerel based fuzzy rugh sets. First, we csider aussia kerel ad prpse aussia kerel based fuzzy rugh sets. Secd we itrduce parameterized attribute reducti with the derived mdel f fuzzy rugh sets. Structures f attribute reducti are ivestigated ad a algrithm with disceribility matrix t fid all reducts is develped. Fially, a heuristic algrithm is desiged t cmpute reducts with aussia kerel fuzzy rugh sets. Several experimets are prvided t demstrate the effectiveess f the idea. Ó 2011 Published by Elsevier Ic. 1. Itrducti As a geeral pursuit i the dmai f machie learig, kerel trick allws mappig data frm iput space it a higher dimesial feature space thrugh kerel fuctis i rder t simplify learig tasks ad make them liear (viz. slvable by liear classifiers [37]). I this way, a umber f liear learig algrithms ca be exteded t deal with liear tasks, such as liear SVMs [42], kerel perceptr [5], kerel discrimiat aalysis [37], liear cmpet aalysis [37], kerel matchig pursuit [43], etc. Mst f them emply feature selecti as a preprcessig step. Accrdig t [34], feature selecti aims at pickig ut sme f the rigial iput features (i) fr perfrmace imprvemet by facilitatig data cllecti ad reducig strage space ad classificati time, (ii) t perfrm sematics aalysis i helpig uderstad the prblem, ad (iii) t imprve predicti accuracy by avidig curse f dimesiality. Accrdig t [2,12,13,24], feature selecti appraches ca be divided it filters [8,9,14,15,28], wrappers [24,44] ad embedded appraches [1,4,51]. Acquirig feedback frm classifiers, the filter methds estimate the classificati perfrmace by sme idirect assessmets, such as distace measures which reflect hw well the classes separate frm each ther. The wrapper methds, the ther had, take the classificati perfrmace f a learig machie as a measure f gdess f a subset f features. Wrapper methds usually prvide mre accurate slutis tha filter methds [26,27,44], but are mre cmputatially expesive. Fially, embedded appraches simultaeusly perfrm feature selecti ad classificati mdelig i the traiig prcess. Crrespdig authr. address: chegdegag@263.et (D. Che) /$ - see frt matter Ó 2011 Published by Elsevier Ic. di: /j.is

2 5170 D. Che et al. / Ifrmati Scieces 181 (2011) Fuzzy rugh sets are exteded frm Pawlak s rugh sets [35] fr dealig with decisi tables with real-valued attributes rather tha symblic es. I the existig framewrk f fuzzy rugh sets a fuzzy similarity relati is emplyed t measure similarity betwee tw bjects. There are tw tpics related with fuzzy rugh sets: develpig differet mdels f fuzzy rugh sets ad perfrmig attribute reducti with fuzzy rugh sets. Sice the pieerig wrk i [6], may effrts [29 31,36,45,46] have bee put the first tpic. Detailed summarizatis mdels f fuzzy rugh sets ca be fud i [48]. O the ther had, attribute reducti with fuzzy rugh sets was first prpsed i [20], where fuzzy depedecy fucti was emplyed t measure gdess f attributes by fuzzy rugh sets prpsed i [6] ad a algrithm t cmpute a reduct was develped. Sme researches attribute reducti with fuzzy rugh sets were maily ctributed t imprve the methd i [20], ad sme key ccepts i the traditial rugh sets, such as cre f reducts, were als geeralized t fuzzy rugh sets [3,16,18 23,41,49,50]. As the rle f fuzzy similarity relati i the framewrk f fuzzy rugh sets, kerels als play the rle as similarity measures i the framewrk f kerel tricks. It was pited ut i [32] that there is a clsed relatiship betwee kerels ad fuzzy similarity relatis, i.e., kerels mappig t the uit iterval with 1 i their diagal are a class f fuzzy similarity relatis. This fact aturally mtivates the idea i this paper t csider such kid f kerels as fuzzy similarity relatis t develp kerel based fuzzy rugh sets, ad attribute reducti with kerel based fuzzy rugh sets ca be prpsed as preprcessig step fr kerel tricks. I this paper we select the well-kw aussia kerels as fuzzy similarity relatis i the framewrk f fuzzy rugh sets. We first develp aussia kerels based fuzzy rugh sets ad discuss their graular structures. Secd we defie parameterized attribute reducti with aussia kerel based fuzzy rugh sets ad develp a algrithm with disceribility matrix t cmpute all the reducts. Here we emply psitive regi i fuzzy rugh sets t measure gdess f subsets f attributes ad attributes are distiguished accrdig t their imprtace related t the decisi. Heuristic algrithm t fid reducts is als prpsed. At last, we emply the reducti algrithm fr aussia kerel SVM as a preprcessig step i classificati learig. Hwever, it is table that every kerel mappig t the uit iterval with 1 i its diagal ca als be csidered as a fuzzy similarity fucti i fuzzy rugh sets, ad differet kerels may have differet techiques t perfrm attribute reducti. We discuss aussia kerels i this paper sice they are widely used i the field f machie learig. The prpsed idea ca be emplyed as a preprcessig step f kerel trick related t aussia kerels. The rest f this paper is rgaized as fllws: Secti 2 itrduces kerel trick ad fuzzy rugh sets. Secti 3 develps attribute reducti with aussia kerel based fuzzy rugh sets; the structure f selected attribute set is characterized by apprach f disceribility matrix i this secti. Experimetal results are described i Secti 4. Cclusis are preseted i Secti Reviews kerels ad fuzzy rugh sets 2.1. Psitive defiite kerels Suppsed X R, H is a Hilbert space. k(x,x 0 ) is a ctiuus ad symmetric fucti X X, if there exists a fucti U : X? H satisfyig that k(x,x 0 )=hu(x),u(x 0 )i H, the k(x,x 0 ) is called a psitive defiite kerel [37]. With a psitive defiite kerel k(x,x 0 ), iput vectrs are mapped it a Hilbert space H, called feature space. Accrdig t [37], kerel trick meas that give a algrithm which is frmulated i terms f a psitive defiite kerel k(x,x 0 ), e ca cstruct a alterative algrithm by replacig k(x,x 0 ) with ather psitive defiite kerel kðx; ~ x 0 Þ. I view f defiiti f psitive defiite kerel, the justificati fr this prcedure is that the rigial algrithm ca be thught f as a dt prduct based algrithm peratig the data U(x 1 ),...,U(x m ). The algrithm btaied by replacig k(x,x 0 ) with kðx; ~ x 0 Þ is the exactly the same dt prduct based algrithm. The ly differece cmes frm that they perate Uðx e 1 Þ;...; Uðx e m Þ. eerally speakig, there are maily tw kids f kerels: traslati ivariat kerels ad dt prduct kerels. The traslati ivariat kerels are idepedet f the abslute psiti f iput x ad x 0. They ly deped the differece betwee x ad x 0. S we have k(x,x 0 )=k(x x 0 ). aussia kerel kðx; x 0 Þ¼exp kx x0 k 2 2r is a well kw traslati ivariat 2 kerel. Sme ther traslati ivariat kerels iclude B -splies kerels, Dirichlet kerels ad Peridic kerels. The secd imprtat family f kerels ca be efficietly described i term f dt prduct, i.e., k(x,x 0 )=k(hx,x 0 i), icludig hmgeeus plymial kerels k(x,x 0 )=hx,x 0 i p ad ihmgeeus plymial kerels k(x,x 0 )=(hx,x 0 i + c) p with c P Fuzzy rugh sets I this subsecti we first review fuzzy lgic peratrs fud i [29,31,36,48], the give a brief itrducti f fuzzy rugh sets. Triagular rms (t-rms fr shrt) have bee rigially studied withi the framewrk f prbabilistic metric spaces [38,39]. I this ctext, t-rms prved t be a apprpriate ccept whe dealig with triagle iequalities. Latter, t-rms ad their dual versi t-crms have bee used t mdel cjucti ad disjucti fr may-valued lgic [7,11,25]. A t-rm is a icreasig, assciative ad cmmutative mappig T : [0,1] [0,1]? [0,1] that satisfies the budary cditi ("x 2 [0,1], T(x,1)= x).

3 D. Che et al. / Ifrmati Scieces 181 (2011) A triagular crm (shrtly t-crm) is a icreasig, assciative ad cmmutative mappig S : [0, 1] [0, 1]? [0, 1] that satisfies the budary cditi ("x 2 [0, 1], S(x, 0) = x). ive a t-rm T, the biary perati # T (a,c) = sup{h 2 [0,1] : T(a,h) 6 c} is called a R-implicatr based T. IfT is lwer semi-ctiuus, the # T is called a residuati implicati f T, rat-residuated implicati. I [29] r is defied by r(a, b) = if{c 2 [0, 1] : S(a, c) P b} as the residuated implicati f a t-crm S. A ifrmati system is a pair A =(U,C), where U ={x 1,...,x } is a empty uiverse f discurse ad C ={a 1,a 2,...,a m } is a empty fiite set f attributes. With a subset f attributes B # C we assciate a biary relati IND(B), called B idisceribility relati defied as IND(B) ={(x, y) 2 U U : a(x) = a(y), "a 2 B}. The IND(B) is a equivalece relati ad IND(B)=\ a2b IND({a}). By [x] B we dete the equivalece class f IND(B) icludig x. Fr X # U the sets [{[x] B : [x] B # X} ad [{[x] B :[x] B \ X /} are called B lwer ad B upper apprximatis f X i A, respectively, deted by BX ad BX. Hwever, the abve traditial rugh set mdel ca just deal with databases described with symblic attributes. This limits the applicatis f rugh sets. Several geeralizatis f the traditial rugh sets were csidered. Amg these geeralizatis, the cmbiati f rugh sets ad fuzzy sets develps a pwerful tl, called fuzzy rugh sets, t deal with real-valued datasets. The defiiti f fuzzy rugh sets was first prpsed i [6]. Sice the may effrts have bee devted t develpig ad characterizig mdels f fuzzy rugh sets. Detailed summaries this tpic ca be fud i [41,45 48]. We here just ffer the basic defiitis f fuzzy rugh sets. Suppse U is a empty uiverse f discurses. As a similarity measure betwee tw bjects, a fuzzy T similarity relati R is a fuzzy set U U which is reflexive, symmetric ad T trasitive, amely R(x,z) P T(R(x,y),R(y,z)) hlds. Fr A 2 F(U), the lwer ad upper apprximatis f A are defied as fllws: (1) T upper apprximati peratr: R T AðxÞ ¼sup u2u TðRðx; uþ; AðuÞÞ; (2) S lwer apprximati peratr: R S A(x) = if u2u S(N(R(x,u)),A(u)); (3) r upper apprximati peratr: R r AðxÞ ¼sup u2u rðnðrðx; uþþ; AðuÞÞ; (4) # lwer apprximati peratr: R # A(x) = if u2u #(R(x,u),A(u)). 3. Apprximatis ad attribute reducti with aussia kerels aussia kerels are widely used i kerel tricks. I this secti we csider aussia kerels as fuzzy T similarity relatis t develp aussia kerel based fuzzy rugh sets ad csider attribute reducti with aussia kerels aussia kerel based fuzzy rugh sets Suppse U ={x 1,x 2,...,x m } is a fiite uiverse f discurses, ad every elemet x i 2 U is described by a vectr (x i1, x i2,...,x i ) 2 R. Thus U is viewed as a subset f R. Sice aussia kerel kðx i ; x j Þ¼exp kx i x j k 2 2r takes values i [0,1], it 2 ca be csidered as a fuzzy relati. We dete this fuzzy relati by R, i.e., R ðx i; x j Þ¼exp kx i x j k2 2r. Obviusly R 2 is reflexive ad symmetric. I [32] it is pited ut that R is T cs trasitive, where T cs ða; bþ ¼ p max ab ffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi 1 a 2 1 b 2 ; 0 is a triagular rm. Thus R is a fuzzy T cs similarity relati. T btai lwer ad upper apprximatis f fuzzy sets related t R, we first derive the residuated implicati f T cs by the fllwig lemma. Lemma ([19,32]). ( 1; a 6 b # Tcs ða; bþ ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ab þ ð1 a 2 Þð1 b 2 Þ; a > b Prf. We have # Tcs ða; bþ ¼supfh 2½0; 1Š : T cs ða; hþ 6 bg, sifa6b, the # Tcs ða; bþ ¼1. Suppse a > b. h shuld satisfy ffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffi ah ð1 a 2 Þð1 h 2 Þ 6 b. It meas ah b 6 ð1 a 2 Þð1 h 2 Þ. ffiffiffiffiffiffiffiffiffiffiffiffiffiffi Let f 1 (h)=ah b ad f 2 ðhþ ¼ ð1 a 2 Þð1 h 2 Þ. The f 1 (h) strictly icreases [0,1], ad f 2 (h) strictly decreases [0,1]. If ffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffi h ¼ ab þ ð1 a 2 Þð1 b 2 Þ, the f 1 (h)=f 2 (h). S if h 6 ab þ ð1 a 2 Þð1 b 2 Þ, the f 1 (h) 6 f 2 (h); if h > ab þ ð1 a 2 Þð1 b 2 Þ, ffiffiffiffiffiffiffiffiffiffiffiffiffiffi the f 1 (h)>f 2 (h). This implies that sup fh 2½0; 1Š : T cs ða; hþ 6 bg ¼ ab þ ð1 a 2 Þð1 b 2 Þ. It shuld be ted that result f Lemma have bee metied i [32] withut prf, here we give a prf f it. h I the rest f this paper we dete # Tcs by # cs fr shrt. With T cs ad # cs aussia kerel based fuzzy rugh sets ca be cmputed as: R AðxÞ ¼sup u2ut cs ðr ðx; uþ; AðuÞÞ; R AðxÞ ¼if u2u# cs ðr ðx; uþ; AðuÞÞ.

4 5172 D. Che et al. / Ifrmati Scieces 181 (2011) The prperties f R ad R are discussed i [48] withi the framewrk f geeral fuzzy rugh sets. Here we ly list the ecessary es fr this paper. Therem [48]. R ad R satisfy the fllwig prperties: (1) R A # A # R A; R A ¼ A () R A ¼ A; (2) R ad R are mte; (3) R ðr AÞ¼R A; R ðr AÞ¼R A; R ðr AÞ¼R A; R ðr AÞ¼R A; (4) R ð[ t2ta t Þ¼[ t2t R A t; R ð\ t2ta t Þ¼\ t2t R A t; (5) If R m # R, the R A # Rm A # A # Rm A # R A. By (1) we kw R A ad R A are a pair f fuzzy sets apprximatig A as upper ad lwer buds, respectively, ad by (5) we ca get that a smaller fuzzy relati ca ffer mre precise apprximatis. These prperties are the theretical fudati f attribute reducti described i Secti 3.2. The fllwig therem shws the graular structure f R A ad R A. Therem R A ¼[fR x k : x k # Ag; R A ¼[ R x k : R x k # A, here x k is a fuzzy set, called fuzzy pit, defied as x k ðyþ ¼ k; y ¼ x. 0; y x Prf. Sice A = [{x k : x k # A}, by (4) f Therem R A ¼[fR x k : x k # Ag is bviusly true. Suppse R A ¼[fR z c : R z c # Ag. Fr every x 2 U, suppse k ¼ R AðxÞ, the we have R x kðyþ ¼T cs ðrðx; yþ; kþ ¼T cs Rðx; yþ; sup R z cðxþ : R z c # A ¼ sup T cs ðrðx; yþ; T cs ðrðz; xþ; cþþ : R z c # A 6 sup T cs ðrðz; yþ; cþ : R z c # A ¼ R AðyÞ: Thus R x k # R A hlds. Ad fr ay k0 > k, clearly R x k 0 cat be icluded by A; therwise R AðxÞ ¼k0. It is a ctradicti. S it implies that R x k is the maximal e i the cllecti fr x g : g 2ð0; 1Šg t be icluded by A. O the ther had, fr u 2 U we have R x k # [ R x b : R x bðuþ 6 AðuÞ. Clearly R x k # \ u2u [ R x b : R x bðuþ 6 AðuÞ. Sice [R x b 2 R x g : g 2ð0; 1Š, we have \ u2u [ R x b : R x bðuþ 6 AðuÞ 2fR x g : g 2ð0; 1Šg; thus R x k ¼\ u2u [ R x b : R x bðuþ 6 AðuÞgÞ. Sice R x k is the maximal e i the cllecti fr x g : g 2ð0; 1Šg icluded by A, it implies k ¼ R x kðxþ ¼if u2u sup fb : T cs ðrðx; uþ; bþ 6 AðuÞg ¼ if u2u # cs ðrðx; uþ; AðuÞÞ. h R Accrdig t Therem 3.1.3, M ¼fR x g : x 2 U; g 2ð0; 1Šg ca be emplyed as the basic graular set t cstruct R ad. This statemet plays a key rle i subsecti 3.2 whe we characterize the structure f reducts Parameterized attribute reducti related t aussia kerel based fuzzy rugh sets Suppse U ={x 1,x 2,...,x m } is a fiite uiverse. Each elemet x i 2 U is described by a set C f attributes with umerical values. The attribute value f x i related t the jth attribute is x ij. The pair (U,C) is a ifrmati system. Suppse U is divided it several disjit parts D 1,D 2,...,D s with a decisi attribute D. The the triple (U,C,D) is called a decisi system. A subset f C iduces a fuzzy T cs similarity relati with the aussia kerel. We dete R ðjþ ðx i; x k Þ¼exp kx ij x kj k 2 2r as 2 the e cmputed with the jth attribute i C, the C ca be equivaletly writte by C ¼ R ð1þ ; Rð2Þ ;...; RðÞ. Firstly, we shuld give the aggregati peratr f multiple elemets i C. I the existig fuzzy rugh sets [3,16 23,41,50] t-rm Mi is used as the aggregati peratr f several fuzzy relatis, ad the fuzzy relati after aggregati is just the itersecti f these fuzzy relatis. Hwever, if we select Mi as the aggregati peratr f elemets i C ¼ R ð1þ ; Rð2Þ ;...; RðÞ, the resultig aggregati des t cicide with R due t the fllwig prperty f aussia kerels: R ðx i; x k Þ¼exp kx i x k k 2 2r ¼ Q 2 s¼1 RðsÞ ðx i; x k Þ. Istead f the t-rm Mi, we itrduce the algebraic prduct T P (x,y)=x y as the aggregati peratr. Clearly, i this case the resultig aggregati f elemets i C ¼ R ð1þ ; Rð2Þ ;...; RðÞ is equal t R. I the fllwig we dete the fuzzy relati aggregated by T P(x,y)=xy with elemets i P # C by R P ad still dete R C by R. Secdly, we shuld develp a methd t measure gdess f subsets f cditial attributes. Fr each D t, t =1,2,...,s, if x R D t, the we have R D tðxþ ¼0. If x 2 D t ; R D tðxþ ¼if u2u # cs ðr ðx; uþ; D tðuþþ ¼ if urdt # cs ðr ðx; uþ; D tðuþþ ¼

5 D. Che et al. / Ifrmati Scieces 181 (2011) ffiffiffiffiffiffiffiffiffiffiffi if urdt 1 ðr ðx; uþþ2. R D tðxþ ca be uderstd as the certaity degree t that x belgs t D t accrdig t the attributes i ffiffiffiffiffiffiffiffiffiffiffi C. Oe bvius bservati is that R D tðxþ is determied by the smallest e (the wrst case) f 1 ðr ðx; uþþ2; u R D t. Thus if there is u 0 R D t such that R ðx; u 0Þ is great eugh, i.e., x is quite similar t a bject i ther classes, the R D tðxþ shuld be very small. Ather bservati is that R D tðxþ ¼1 is ever true because R ðx; uþ 0 always hlds, i.e., every pair f bjects are similar i a certai degree with respect t aussia kerels. Ps C ðdþ ¼[ s t¼1 R D t is called the psitive regi f decisi attribute D related t the cditial attribute set C, we will emply psitive regi as a measure f gdess f attributes. Hwever, if we use aussia fucti as the similarity fucti, deletig ay attribute C frm C will result i Ps C D(x i )> Ps C {C} D(x i ) fr every i =1,2,...,m. S we cat emply the idea i the traditial rugh sets [35,40] ad existig fuzzy rugh sets [41] t defie attribute reduct as the miimal subset f C t keep the psitive regi ivariat. This issue is als t metied i [19]. We vercme this prblem by csiderig a threshld e f the psitive regi, ad we ca defie a parameterized attribute reduct with aussia kerel based fuzzy rugh sets by limitig the chage f psitive regi withi the give threshld e. The idea ca be frmulated as fllws. Defiiti Suppse (U,C,D) is a decisi system, e 2 [0,1]. Fr C 2 C, if Ps C D(x i ) Ps C {C} D(x i ) 6 e fr every i =1,2,...,m, the C is called e superfluus i C relative t D; therwise C is called e idispesable i C relative t D. Fr every P # C, ifps C D(x i ) Ps P D(x i ) 6 e fr every i =1,2,...,m, ad every elemet i P is idispesable, the P is called a e reduct f C relative t D. The cllecti f all the e idispesable elemets i C is called the e cre f C relative t D, deted by Cre D (C), ad we have the fllwig therem fr the cre. Therem Cre D (C)=\Red D (C), where Red D (C) is the cllecti f all the e reduct f C relative t D. Prf. If C is e idispesable i C relative t D, the C shuld be icluded i every e reduct f C. Hece Cre D (C) # \Red D (C). O the ther had, if C is e superfluus i C relative t D, the C {C}ctais a e reduct f C, thus there is a e reduct f C that des t iclude C, hece C R \Red D (C). It implies Cre D (C) \Red D (C). h Fr C 2 C, clearly C is e superfluus i C relative t D if ad ly if R D tðxþ e 6 R C fcg D t ðxþ fr every D t, t =1,2,...,l ad x 2 D t, ad if ad ly if x kðxþ # R C fcg D t fr x 2 D t ad kðxþ ¼R D tðxþ e, ad if ad ly if R C fcg x kðxþ # R C fcg D t fr x 2 D t by Therem 3.1.3, here x k(x) is a fuzzy pit. Thus we have the fllwig therem. Therem Suppse P # C. P ctais a e reduct f C if ad ly if R P x kðxþðzþ ¼0 fr x 2 D t,zr D t,t=1,2,...,l. Prf. P ctais a e reduct f C if ad ly if R P x kðxþ # R P D t fr x 2 D t.ifr P x kðxþ # R P D t, the clearly R P x kðxþðzþ ¼0 fr z R D t. h Cversely, if R P x kðxþðzþ ¼0 fr z R D t, the R P x kðxþ # D t which implies R P x kðxþ # R P D t by Therem qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Therem Suppse P # C. P ctais a e reduct f C if ad ly if there is Q # P such that R Q ðx; zþ 6 1 k 2 ðxþ fr x 2 D t,zrd t,t=1,2,...,l. Prf. R P x kðxþðzþ ¼0 fr x 2 D t ; z R D t ; t ¼ 1; 2;...; l () sup u2u T cs ðr P ðz; uþ; x kðxþðuþþ ¼ 0 () T cs ðr P ðx; zþ; kðxþþ ¼ 0 () RP ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðx; zþkðxþ ð1 ðr P ðx; zþþ2 Þð1 k 2 ðxþþ 6 0 () R P ðx; zþ 6 1 k 2 ðxþ () there is Q # P such that R Q ðx; zþ 6 1 k 2 ðxþ. By Therem we fiish the prf. h Therem will be applied t study the structure f e reducti f C ad desig algrithms t cmpute all the e reducts f C i the fllwig subsecti Disceribility matrix based attribute reducti Disceribility matrix is a key ccept t ivestigate attribute reducti i the rugh set framewrk [40]. A reasable defiiti f disceribility matrix ca reveal the structure f attribute reducti, furthermre, it is the theretical fudati t desig algrithms t cmpute reducts. I this subsecti we develp a apprach t fid the e reducts based disceribility matrix. Defiiti Suppse (U,C,D) is a decisi system. By M(U,C,D) we dete a m m matrix (c ij ) mm, called the disceribility matrix f (U, C, D), defied as

6 5174 D. Che et al. / Ifrmati Scieces 181 (2011) (1) if x i ad x j belg t differet decisi classes, c ij ={^P : P # C}, here ^P is the cjucti f elemets i P, ad P satisfies R P ðx i; x j Þ 6 1 k 2 ðx i Þ ad fr Q # P such that R Q ðx i; x j Þ 6 1 k 2 ðx i Þ, the Q = P. (2) c ij = /, therwise. Clearly c ij is the cllecti f all the cjuctis f elemets i P # C thus that P is a miimal e satisfyig R P ðx i; x j Þ 6 1 k 2 ðx i Þ. It is remarkable that M(U,C,D) may t be symmetric i this case. Therem Suppse (U,C,D) is a decisi system, P # C. We have the fllwig tw statemets: (1) P ctais a e reduct f C if ad ly if P \ c ij / fr c ij /, here P \ c ij defied as [{Q # P : ^Q 2 c ij }. (2) Cre D (C)=[{Q ij # C : Q ij = \{P : ^P 2 c ij },i,j=1,2,...,m}. Prf (1) If P ctais a e reduct f C, the fr x i ad x j belg t differet decisi classes, there exists a miimal Q # P such that R Q ðx i; x j Þ 6 1 k 2 ðx i Þ, thus ^Q 2 c ij ad P \ c ij /. Cversely, if P \ c ij / fr c ij /, the there exists a miimal Q # P such that R Q ðx i; x j Þ 6 1 k 2 ðx i Þ, thus P ctais a e reduct f C. (2) If C 2 Cre D (C), the there exist x i ad x j belgig t differet decisi classes such that R C fcg ðx i ; x j Þ > 1 k 2 ðx i Þ.S if P # C such that R P ðx i; x j Þ 6 1 k 2 ðx i Þ, the C 2 P must hld. This implies C 2\fP : ^P 2 c ij g ad Cre D ðcþ # [ Q ij # C : Q ij ¼\fP : ^P 2 c ij g; i; j ¼ 1; 2;...; m : Cversely, if C 2[{Q ij # C:Q ij = \{P : ^P 2 c ij }, i,j =1,2,...,m}, the there exist x i ad x j belgig t differet decisi classes such that C 2\{P : ^P 2 c ij }, s R C fcg ðx i ; x j Þ > 1 k 2 ðx i Þ hlds, which implies C 2 Cre D (C). Thus we fiish the prf. h (2) f Therem prpses a frmula t cmpute the relative cre by disceribility matrix, this frmula will play a key rle whe desig algrithm t cmpute e reduct i Secti 4.1. Crllary Suppse (U,C,D) is a decisi system, P # C. P is a e reduct f C if ad ly if P is the miimal subset f C satisfyig P \ c ij / fr c ij /. A disceribility fucti f(u,c,d) fr (U,C,D) is a Blea fucti f Blea variables C 1 ; C 2 ;...; C crrespdig t the attributes C 1,C 2,...,C i C, respectively, ad defied as f ðu; C; DÞðC 1 ; C 2 ;...; C Þ ¼ ^f_ðc ij Þ : c ij /; 1 6 i; j 6 mg, where _(c ij ) is the disjucti f all elemets i c ij as ^P. By usig f the disceribility fucti, we have the fllwig therem t cmpute all the e reducts f C. Therem Suppse (U,C,D) is a decisi system; M(U,C,D)= (c ij :i,j6 ) is the disceribility matrix f (U,C,D) ad f(u,c,d) is the disceribility fucti f (U,C,D). If f ðu; C; DÞ ¼_ l k¼1 ð^d kþðd k # CÞ is cmputed frm f(u,c,d) by applyig the multiplicati ad absrpti laws as may times as pssible such that every elemet i D i ly appears e time, the the set {D k :k6 l} is the cllecti f all the e reducts f C, i.e., Red D (C)={D 1,...,D l }. Prf. Fr every k =1,...,l, we have D k \ c ij /. Sice f ðu; C; DÞ ¼_ l ð^d k¼1 kþ, fr every D k, if we reduce a elemet C i D k ðd 0 k ¼ D k fcgþ, the f ðu; C; DÞ _ r¼1 k 1ð^D rþ_ð^d 0 k Þ_ð_l D r¼kþ1 rþ ad f ðu; C; DÞ < _ r¼1 k 1ð^D rþ_ð^d 0 k Þ_ð_l D r¼kþ1 rþ. If"c ij,we have D 0 k \ c ij /, the ^D 0 k 6 _c ij, which implies f ðu; C; DÞ P k 1 _ ð^d r Þ_ð^D 0 k Þ l D r ad f ðu; C; DÞ ¼ k 1 _ ð^d r Þ_ð^D 0 k Þ l D r r¼1 r¼kþ1 r¼1 r¼kþ1 it is a ctradicti. Hece there exists c i0 j 0 such that D 0 k \ c i 0 j 0 ¼ /, which implies D k is a reducti f (U,C,D). Fr every X 2 Red D (C), we have X \ c ij / fr every c ij /. S we have f(u,c,d) ^ (^X)=^(_c ij ) ^ (^X)=^X. This implies ^X 6 f(u,d,d). Suppse that fr every k we have D k X /. The fr every k e ca fid C k 2 D k X. By rewritig f ðu; C; DÞ ¼ð_ l k¼1 C kþ^u, we have ^X 6 _ l k¼1 C k. S there is C k0, such that ^X 6 C k0. This implies C k0 2 X, which is a ctradicti. S D k0 # X fr sme k 0, sice bth X ad D k0 are reducts. We have X ¼ D k0. Hece Red D (C)={D 1,...,D l }. h Nw we ca cclude that C ca be categrized it three parts accrdig t their imprtace related t the classificati: (1) elemets i the cre f reducts which shuld be icluded i every reduct; (2) elemets cat be icluded i ay reduct; (3) elemets belg t sme but t all reducts. This partiti als seems reasable i the practical viewpit.

7 D. Che et al. / Ifrmati Scieces 181 (2011) It is wrth pitig ut that the prpsed idea i this paper is t ly limited t aussia kerels, but als applicable t all kerels mappig t the uit iterval with 1 i its diagal. We ca develp fuzzy rugh sets ad csider attribute reducti with this kid f kerels. Hwever, differet kerels may have differet techiques t perfrm attribute reducti. Fr example, we emply the algebraic prduct T P (x,y)=x y as the aggregati peratr fr aussia kerels, ad i the existig attribute reducts [3,10,16 23,41,50] t-rm Mi is emplyed as aggregati peratr fr fuzzy Mi similarity relatis, this differece may lead t differet frmulati f disceribility matrixes. I additi, sice a kerel plays the same rle as a similarity measure i bth attribute reducti ad kerel trick, we suggest t use the same kerel whe attribute reducti is emplyed as a preprcessig step f kerel trick. 4. Experimets ad cmpariss I this secti we will desig a algrithm t cmpute reducts; we will als perfrm attribute reducti as a preprcessig step fr aussia kerel supprt vectr machies i rder t test the effectiveess f the prpsed wrk. Table 1 Descripti f experimetal data. Data Samples Features Classes 1 Credit Heart Hepatitis Hrse I Sar Wie Wpbc (a) umber f selected featues (b) classificati perfrmace f selected features Fig. 1. Variati f size f selected features ad crrespdig classificati perfrmace (sar) (a) umber f selected features (b) classificati accuracies f selected features Fig. 2. Variati f size f selected features ad crrespdig classificati perfrmace (wie).

8 5176 D. Che et al. / Ifrmati Scieces 181 (2011) Algrithm desig ad cmplexity aalysis The algrithm by disceribility matrix is helpful t fid all the reducts f the dataset, but the time cmplexity t fid all the reducts icreases expetially with the umber f attributes O(jUj 2 2 jcj ) [40], where juj is the size f the uiverse, jcj is the umber f cditial attributes. I real applicatis, it is t ecessary t fid all the reducts. It is eugh t address the real prblem by usig e f the reducts. I the fllwig we prvide a heuristic algrithm t fid a reduct. Iput: (U,C,D), Reduct {} Step 1: Cmpute the similarity relati f the set f all cditi attributes: R ; Step 2: Cmpute Ps C ðdþ ¼[ s t¼1 R D t; Step 3: Cmpute c ij by its defiiti i Secti 3; Step 4: Cmpute Cre D (C)=[ {Q ij # C : Q ij = \{P : ^P 2 c ij }, i,j =1,2,...,m}; Delete thse c ij with empty verlap withcre D (C); Step 5: Let Reduct = Cre D (C); Step 6: Add the elemet a whse frequecy f ccurrece is maximum i all c ij it Reduct; ad delete thse c ij with empty verlap with Reduct; Step 7: If there still exist sme c ij /, g t Step 6; Otherwise, g t Step 8; Step 8: If Reduct is t idepedet, delete the redudat elemets i Reduct; Step 9: Output Reduct. The cmputatial cmplexity f this algrithm is O(jUj 2 jcj) Experimetal aalysis I this subsecti, we will perfrm experimets t examie effectiveess f ur idea. We select aussia kerel SVM as a classifier t validate the quality f the features selected by ur techique. Eight datasets are dwladed frm UCI machie learig repsitry [33], described i Table 1. First, we csider the impact f parameters feature selecti. We set r frm 0.1 t 0.4 with step I the meawhile, e is set as 0.01 t 0.1 with step With these parameters, we ca get 100 subsets f attributes ad the crrespdig classificati perfrmace. We perfrm experimets data sets sar ad wie. The results are shw i Figs. 1 ad 2, where the x-axis is r ad y-axis ise. As the bjective f feature selecti is t fid a miimal subspace which has gd classificati perfrmace, s it is expected that the size f the selected feature is relatively small ad the crrespdig classificati perfrmace is gd eugh. We ca see frm the abve results that [0.1,0.2] ad [0.01,0.02] are prper value dmais fr r ad e, respectively. Table 2 Numbers f selected features. Data Raw data KFRS CFS NRS RS Credit Heart Hepatitis Hrse I Sar Wie Wpbc Table 3 Classificati accuracies based aussia kerel SVM (%). Data Raw data KFRS CFS NRS RS Credit ± ± ± ± ± 18.5 Heart ± ± ± ± 6.59 Hepatitis ± ± ± ± ± 7.24 Hrse ± ± ± ± ± 4.45 I ± ± ± ± ± 5.97 Sar ± ± ± ± 7.60 Wie ± ± ± ± ± 4.10 Wpbc ± ± ± ± ± 5.06

9 D. Che et al. / Ifrmati Scieces 181 (2011) We ca see that the classificati accuracies f the reduced data are relatively high ad the sizes f the reduced data are small if we let r ad e take values i these dmais, respectively. Nw we cmpare the umber f the selected features ad classificati perfrmaces f the reducts, shw i Tables 2 ad 3, where reduct is cmputed by the prpsed algrithm ad the classificati perfrmaces f reducts are attaied with aussia kerel SVM based the 10-fld crss validati techique ad aussia kerel SVM is implemeted with su_svm3.00 tlbx. d ad e are specified as 0.1 ad 0.02, respectively. Nw we aalyze experimetal results i Tables 2 ad 3. Cmpared with the raw data, we see that (i) amg the 8 data sets, ur prpsed attribute reducti methd perfrms well six data sets: hepatitis, hrse, wpbc, aeal, i, sar ad wie. Fr these six data sets, umbers f attributes greatly decrease after reducti cmpared with the raw data, ad perfrmaces f classifier (SVM) are imprved distictly. This implies ur prpsed attribute reducti methd ca really delete redudat attributes frm these data sets; (ii) fr data sets credit ad heart, few attributes are deleted, ad imprvemets f perfrmaces f classifier are t sigificat. Hwever, this may due t that there are less redudat attributes i these tw data sets sice the rigial umbers f attributes i these tw data sets are few. I rder t cmpare the prpsed techiques with the existig e, we use eighbrhd rugh set apprach (NRS) [18] ad crrelati based feature selecti (CFS) [14] these data sets. These techiques ca deal with umerical features directly. Frm Tables 2 ad 3, we ca als see that fuzzy rugh sets are better tha ther algrithms imst cases. I additi, we als itrduce the classical rugh set techique t select features with a frward greedy search strategy, deted by RS. As t datasets heart ad sar, feature is retured. This pheme has bee metied i the previus wrk as ay sigle feature prduces the depedecy f zer. S the algrithm stps here. I additi, we gather six cacer recgiti tasks utlied i Table 4. The umbers f features are much mre tha the umbers f samples i these tasks. The detailed descripti abut these tasks ca be gtte frm the webpage ( Overfittig is the mst imprtat challege i gee classificati. Attribute reducti may help vercme this prblem. We perfrm attribute reducti based techiques f eighbrhd rugh sets ad fuzzy rugh sets, respectively. The results are preseted i Tables 5 ad 6. We see that mst f cadidate gees are remved frm classificati learig ad ly a few gees are selected. Mrever, the gees selected by FRS are a little mre tha that by NRS; hwever; the classificati perfrmace is greatly imprved by FRS cmpared with the raw data ad thse selected by NRS. These results shw fuzzy rugh sets are useful i gee selecti fr cacer recgiti. Table 4 ee expressi data sets. Data ees Class Samples Leuk Leuk2 12, SRBCT Breast Lug2 12, DLBCL Table 5 Number f the features selected. Data Raw NRS FRS Breast DLBCL Leukemial Leukemial Lug SRBCT Table 6 Accuracy f the selected gees. Data Raw (%) NRS (%) FRS (%) Breast DLBCL Leukemial Leukemial Lug SRBCT

10 5178 D. Che et al. / Ifrmati Scieces 181 (2011) Cclusi ad future wrk Fuzzy rugh sets are a ht tpic i graular cmputig. I this paper we itrduce aussia kerel it fuzzy rugh sets fr cmputig fuzzy similarity relati ad develp a vel methd f attribute reducti with parameter based the prpsed mdel. We discuss the structure f subsets f selected attributes with fuzzy disceribility matrix. Attributes ca be gruped as three cllectis accrdig t their imprtace related t the decisi. The mai purpse f this paper is t develp attribute reducti with kerel tricks. We use the UCI machie learig data sets ad cacer classificati tasks t test the prpsed techique. The experimetal results shws aussia kerel based fuzzy rugh sets ca fid gd subsets f attributes fr classificati learig. Althugh aussia kerel is frequetly used, there are als sme ther kerel fuctis ca be itrduced it fuzzy rugh sets. We will wrk ther kerels ad develp a set f attribute reducti techiques based fuzzy rugh sets ad kerels. Ackwledgemets This paper is partly supprted by Natial Natural Sciece Fudati uder rats , , ad ad a grat f Natial Basic Research Prgram f Chia (2009CB ). Refereces [1] M.F. Balca, A. Blum, S. Vempala, Kerels as features: kerels, margis, ad lw-dimesial mappigs, Machie Learig 65 (2006) [2] O. Barzilay, V.L. Brailvsky, O dmai kwledge ad feature selecti usig a supprt vectr machie, Patter Recgiti Letters 20 (1999) [3] R.B. Bhatt, M. pal, O fuzzy rugh sets apprach t feature selecti, Patter recgiti Letters 26 (2005) [4] P.S. Bradley, O.L. Magasaria, Feature selecti via ccave miimizati ad supprt vectr machie, i: Prceedigs f the 15th Iteratial Cferece Machie Learig, Sa Fracisc, CA, USA, 1998, pp [5] J.H. Che, C.S. Che, Fuzzy kerel perceptr, IEEE Trasactis Neural Netwrks 13 (2002) [6] D. Dubis, H. Prade, Rugh fuzzy sets ad fuzzy rugh sets, Iteratial Jural f eeral Systems 17 (1990) [7] D. Dubis, H. Prade, A review f fuzzy set aggregati cectives, Ifrmati Scieces 36 (1985) [8] R. Duda, P. Hart, D. Strk, Patter Classificati, secd ed., Jh Wiley & Ss, New Yrk, NY, USA, [9] T. Evgeiu, M. Ptil, C. Papagergiu, T. Pggi, Image represetatis ad feature selecti fr multimedia database search, IEEE Trasactis Kwledge ad Data Egieerig 15 (2003) [10] S. Feradez, J.M. Murakami, Rugh set aalysis f a geeral type f fuzzy data usig trasitive aggregatis f fuzzy similarity relatis, Fuzzy Sets ad Systems 139 (2003) [11] S. ttwald, Fuzzy Sets ad Fuzzy Lgic, Vieweg, Brauschweig, [12] I. uy, A. Elisseeff, A itrducti t variable ad feature selecti, Jural f Machie Learig Research 3 (2003) [13] I. uy, J. West, S. Barhill, V. Vapik, ee selecti fr cacer classificati usig supprt vectr machies, Machie Learig 46 (2002) [14] M. Hall, Crrelati-based feature selecti fr discrete ad umeric class machie learig, i: Prceedigs f the 17th ICML, CA, 2000, pp [15] C.L. Huag, C.J. Wag, A A-based feature selecti ad parameters ptimizati fr supprt vectr machies, Expert Systems with Applicatis 31 (2006) [16] Q.H. Hu, D.R. Yu, Z.X. Xie, Ifrmati-preservig hybrid data reducti based fuzzy-rugh techiques, Patter Recgiti Letters 27 (2006) [17] Q.H. Hu, Z.X. Xie, D.R. Yu, Hybrid attribute reducti based a vel fuzzy-rugh mdel ad ifrmati graulati, Patter Recgiti 40 (2007) [18] Q.H. Hu, D.R. Yu, J. F Liu, C. X Wu, Neighbrhd rugh set based hetergeeus feature subset selecti, Ifrmati Scieces 178 (2008) [19] Q.H. Hu, L. Zhag, D.. Che, W. Pedrycz, D. Yu, aussia kerel based fuzzy rugh sets: mdel, ucertaity measures ad applicatis, Iteratial Jural f Apprximatig Reasig 51 (2010) [20] R. Jese, Q. She, Fuzzy-rugh attributes reducti with applicati t web categrizati, Fuzzy Sets ad Systems 141 (2004) [21] R. Jese, Q. She, Fuzzy-rugh sets assisted attribute selecti, IEEE Trasactis Fuzzy Systems 15 (2007) [22] R. Jese, Q. She, Sematics-preservig dimesiality reducti: rugh ad fuzzy rugh based appraches, IEEE Trasactis Kwledge ad Data Egieerig 16 (2004) [23] R. Jese, Q. She, New appraches t fuzzy-rugh feature selecti, IEEE Trasactis Fuzzy Systems 17 (2009) [24].H. Jh, R. Khavi, K. Pfleger, Irrelevat features ad the subset selecti prblem, i: Prceedigs f the 11th Iteratial Cferece Machie Learig, 1994, pp [25] E.P. Klemet, R. Mesiar, E. Pap, Triagular rms, Treds i Lgic, vl. 8, Kluwer Academic Publishers, Drdrecht, [26] J. Khavi, Wrappers fr feature subset selecti, AIJ issue relevace, 1995 [27] Y. Liu, Y.F. Zheg, FS-SFS: a vel feature selecti methd fr supprt vectr machies, Patter Recgiti 39 (2006) [28] K.Z. Ma, Feature subset selecti fr supprt vectr machies thrugh discrimiative fucti pruig aalysis, IEEE Trasactis Systems, Ma, ad Cyberetics Part B: Cyberetics 34 (2004) [29] J.S. Mi, W.X. Zhag, A aximatic characterizati f a fuzzy geeralizati f rugh sets, Ifrmati Scieces 160 (2004) [30] J.S. Mi, Y. Leug, H.Y. Zha, eeralized fuzzy rugh sets determied by a triagular rm, Ifrmati Scieces 178 (2008) [31] N.N. Mrsi, M.M. Yakut, Aximatics fr fuzzy rugh sets, Fuzzy Sets ad Systems 100 (1998) [32] B. Mser, O represetig ad geeratig kerels by fuzzy equivalece relatis, Jural f Machie Learig Research 7 (2006) [33] D.J. Newma, S. Hettich, C.L. Blake, C.J. Merz, UCI Repsitry f machie learig databases, Uiversity f Califria, Departmet f Ifrmati ad Cmputer Sciece, Irvie, CA, < [34] J. Neuma, C. Schrr,. Steidl, Cmbied SVM-based feature selecti ad classificati, Machie Learig 61 (2005) [35] Z. Pawlak, Rugh sets, Iteratial Jural f Cmputer Ifrmati Sciece 11 (1982) [36] A.M. Radzikwska, E.E. Kerre, A cmparative study f fuzzy rugh sets, Fuzzy Sets ad Systems 126 (2002) [37] B. Schlkpf, A.J. Smla, Learig with Kerels, The MIT Press, [38] B. Schweizer, A. Sklar, Assciative fuctis ad statistical triagle iequalities, Publicaties Mathematicae Debrece 8 (1961) [39] B. Schweizer, A. Sklar, Prbabilistic Metric Spaces, Nrth-Hllad, Amsterdam, 1983.

11 D. Che et al. / Ifrmati Scieces 181 (2011) [40] A. Skwr, C. Rauszer, The disceribility matrices ad fuctis i ifrmati systems, i: R. Slwiski (Ed.), Itelliget Decisi supprt, Hadbk f Applicatis ad Advaces f the Rugh Sets Thery, Kluwer Academic Publishers, [41] C.C.E. Tsag, D.. Che, S.D. Yueg, W.T.J. Lee, X.Z. Wag, Attribute reducti usig fuzzy rugh sets, IEEE Trasactis Fuzzy Systems 16 (2008) [42] V.N. Vapik, The Nature f Statistical Learig Thery, Spriger, New Yrk, [43] P. Vicet, Y. Begi, Kerel matchig pursuit, Machie Learig 48 (2002) [44] J. West, S. Mukherjee, O. Chapelle, M. Ptil, T. Pggi, V. Vapik, Feature selecti fr SVMs, i: T.K. Lee, T.. Dietterich, V. Tresp (Eds.), Advaces i Neural Ifrmati Prcessig Systems, vl. 13, MIT Press, Cambridge, MA, USA, 2001, pp [45] W.Z. Wu, W.X. Zhag, Cstructive ad aximatic appraches f fuzzy apprximati peratrs, Ifrmati Scieces 159 (2004) [46] W.Z. Wu, J.S. Mi, W.X. Zhag, eeralized fuzzy rugh sets, Ifrmati Scieces 151 (2003) [47] W.Z. Wu, Attribute reducti based evidece thery i icmplete decisi systems, Ifrmati Scieces 178 (2008) [48] S.D. Yeug, D.. Che, C.C.E. Tsag, W.T.J. Lee, X.Z. Wag, O the geeralizati f fuzzy rugh sets, IEEE Trasactis Fuzzy Systems 13 (2005) [49] D.R. Yu, Q.H. Hu, C.X. Wu, Ucertaity measures fuzzy relatis ad their applicatis, Applied Sft Cmputig 7 (2007) [50] S.Y. Zha, E.C.C. Tsag, O fuzzy apprximati peratrs i attribute reducti with fuzzy rugh sets, Ifrmati Scieces 178 (2008) [51] J. Zhu, S. Rsset, T. Hastie, R. Tibshirai, 1-Nrm supprt vectr machies, i: S. Thru, L. Saul, B. Schlkpf (Eds.), Advaces i Neural Ifrmati Prcessig Systems, vl. 13, MIT Press, Cambridge, MA, USA, 2004.

Multi-objective Programming Approach for. Fuzzy Linear Programming Problems

Multi-objective Programming Approach for. Fuzzy Linear Programming Problems Applied Mathematical Scieces Vl. 7 03. 37 8-87 HIKARI Ltd www.m-hikari.cm Multi-bective Prgrammig Apprach fr Fuzzy Liear Prgrammig Prblems P. Padia Departmet f Mathematics Schl f Advaced Scieces VIT Uiversity

More information

Author. Introduction. Author. o Asmir Tobudic. ISE 599 Computational Modeling of Expressive Performance

Author. Introduction. Author. o Asmir Tobudic. ISE 599 Computational Modeling of Expressive Performance ISE 599 Cmputatial Mdelig f Expressive Perfrmace Playig Mzart by Aalgy: Learig Multi-level Timig ad Dyamics Strategies by Gerhard Widmer ad Asmir Tbudic Preseted by Tsug-Ha (Rbert) Chiag April 5, 2006

More information

ENGI 4421 Central Limit Theorem Page Central Limit Theorem [Navidi, section 4.11; Devore sections ]

ENGI 4421 Central Limit Theorem Page Central Limit Theorem [Navidi, section 4.11; Devore sections ] ENGI 441 Cetral Limit Therem Page 11-01 Cetral Limit Therem [Navidi, secti 4.11; Devre sectis 5.3-5.4] If X i is t rmally distributed, but E X i, V X i ad is large (apprximately 30 r mre), the, t a gd

More information

A New Method for Finding an Optimal Solution. of Fully Interval Integer Transportation Problems

A New Method for Finding an Optimal Solution. of Fully Interval Integer Transportation Problems Applied Matheatical Scieces, Vl. 4, 200,. 37, 89-830 A New Methd fr Fidig a Optial Sluti f Fully Iterval Iteger Trasprtati Prbles P. Padia ad G. Nataraja Departet f Matheatics, Schl f Advaced Scieces,

More information

Intermediate Division Solutions

Intermediate Division Solutions Itermediate Divisi Slutis 1. Cmpute the largest 4-digit umber f the frm ABBA which is exactly divisible by 7. Sluti ABBA 1000A + 100B +10B+A 1001A + 110B 1001 is divisible by 7 (1001 7 143), s 1001A is

More information

5.1 Two-Step Conditional Density Estimator

5.1 Two-Step Conditional Density Estimator 5.1 Tw-Step Cditial Desity Estimatr We ca write y = g(x) + e where g(x) is the cditial mea fucti ad e is the regressi errr. Let f e (e j x) be the cditial desity f e give X = x: The the cditial desity

More information

Ch. 1 Introduction to Estimation 1/15

Ch. 1 Introduction to Estimation 1/15 Ch. Itrducti t stimati /5 ample stimati Prblem: DSB R S f M f s f f f ; f, φ m tcsπf t + φ t f lectrics dds ise wt usually white BPF & mp t s t + w t st. lg. f & φ X udi mp cs π f + φ t Oscillatr w/ f

More information

Study of Energy Eigenvalues of Three Dimensional. Quantum Wires with Variable Cross Section

Study of Energy Eigenvalues of Three Dimensional. Quantum Wires with Variable Cross Section Adv. Studies Ther. Phys. Vl. 3 009. 5 3-0 Study f Eergy Eigevalues f Three Dimesial Quatum Wires with Variale Crss Secti M.. Sltai Erde Msa Departmet f physics Islamic Aad Uiversity Share-ey rach Ira alrevahidi@yah.cm

More information

Fourier Method for Solving Transportation. Problems with Mixed Constraints

Fourier Method for Solving Transportation. Problems with Mixed Constraints It. J. Ctemp. Math. Scieces, Vl. 5, 200,. 28, 385-395 Furier Methd fr Slvig Trasprtati Prblems with Mixed Cstraits P. Padia ad G. Nataraja Departmet f Mathematics, Schl f Advaced Scieces V I T Uiversity,

More information

D.S.G. POLLOCK: TOPICS IN TIME-SERIES ANALYSIS STATISTICAL FOURIER ANALYSIS

D.S.G. POLLOCK: TOPICS IN TIME-SERIES ANALYSIS STATISTICAL FOURIER ANALYSIS STATISTICAL FOURIER ANALYSIS The Furier Represetati f a Sequece Accrdig t the basic result f Furier aalysis, it is always pssible t apprximate a arbitrary aalytic fucti defied ver a fiite iterval f the

More information

ENGI 4421 Central Limit Theorem Page Central Limit Theorem [Navidi, section 4.11; Devore sections ]

ENGI 4421 Central Limit Theorem Page Central Limit Theorem [Navidi, section 4.11; Devore sections ] ENGI 441 Cetral Limit Therem Page 11-01 Cetral Limit Therem [Navidi, secti 4.11; Devre sectis 5.3-5.4] If X i is t rmally distributed, but E X i, V X i ad is large (apprximately 30 r mre), the, t a gd

More information

Mean residual life of coherent systems consisting of multiple types of dependent components

Mean residual life of coherent systems consisting of multiple types of dependent components Mea residual life f cheret systems csistig f multiple types f depedet cmpets Serka Eryilmaz, Frak P.A. Cle y ad Tahai Cle-Maturi z February 20, 208 Abstract Mea residual life is a useful dyamic characteristic

More information

Chapter 3.1: Polynomial Functions

Chapter 3.1: Polynomial Functions Ntes 3.1: Ply Fucs Chapter 3.1: Plymial Fuctis I Algebra I ad Algebra II, yu ecutered sme very famus plymial fuctis. I this secti, yu will meet may ther members f the plymial family, what sets them apart

More information

UNIVERSITY OF TECHNOLOGY. Department of Mathematics PROBABILITY THEORY, STATISTICS AND OPERATIONS RESEARCH GROUP. Memorandum COSOR 76-10

UNIVERSITY OF TECHNOLOGY. Department of Mathematics PROBABILITY THEORY, STATISTICS AND OPERATIONS RESEARCH GROUP. Memorandum COSOR 76-10 EI~~HOVEN UNIVERSITY OF TECHNOLOGY Departmet f Mathematics PROBABILITY THEORY, STATISTICS AND OPERATIONS RESEARCH GROUP Memradum COSOR 76-10 O a class f embedded Markv prcesses ad recurrece by F.H. Sims

More information

A Study on Estimation of Lifetime Distribution with Covariates Under Misspecification

A Study on Estimation of Lifetime Distribution with Covariates Under Misspecification Prceedigs f the Wrld Cgress Egieerig ad Cmputer Sciece 2015 Vl II, Octber 21-23, 2015, Sa Fracisc, USA A Study Estimati f Lifetime Distributi with Cvariates Uder Misspecificati Masahir Ykyama, Member,

More information

IJISET - International Journal of Innovative Science, Engineering & Technology, Vol. 2 Issue 12, December

IJISET - International Journal of Innovative Science, Engineering & Technology, Vol. 2 Issue 12, December IJISET - Iteratial Jural f Ivative Sciece, Egieerig & Techlgy, Vl Issue, December 5 wwwijisetcm ISSN 48 7968 Psirmal ad * Pararmal mpsiti Operatrs the Fc Space Abstract Dr N Sivamai Departmet f athematics,

More information

Fourier Series & Fourier Transforms

Fourier Series & Fourier Transforms Experimet 1 Furier Series & Furier Trasfrms MATLAB Simulati Objectives Furier aalysis plays a imprtat rle i cmmuicati thery. The mai bjectives f this experimet are: 1) T gai a gd uderstadig ad practice

More information

BIO752: Advanced Methods in Biostatistics, II TERM 2, 2010 T. A. Louis. BIO 752: MIDTERM EXAMINATION: ANSWERS 30 November 2010

BIO752: Advanced Methods in Biostatistics, II TERM 2, 2010 T. A. Louis. BIO 752: MIDTERM EXAMINATION: ANSWERS 30 November 2010 BIO752: Advaced Methds i Bistatistics, II TERM 2, 2010 T. A. Luis BIO 752: MIDTERM EXAMINATION: ANSWERS 30 Nvember 2010 Questi #1 (15 pits): Let X ad Y be radm variables with a jit distributi ad assume

More information

K [f(t)] 2 [ (st) /2 K A GENERALIZED MEIJER TRANSFORMATION. Ku(z) ()x) t -)-I e. K(z) r( + ) () (t 2 I) -1/2 e -zt dt, G. L. N. RAO L.

K [f(t)] 2 [ (st) /2 K A GENERALIZED MEIJER TRANSFORMATION. Ku(z) ()x) t -)-I e. K(z) r( + ) () (t 2 I) -1/2 e -zt dt, G. L. N. RAO L. Iterat. J. Math. & Math. Scl. Vl. 8 N. 2 (1985) 359-365 359 A GENERALIZED MEIJER TRANSFORMATION G. L. N. RAO Departmet f Mathematics Jamshedpur C-perative Cllege f the Rachi Uiversity Jamshedpur, Idia

More information

Solutions. Definitions pertaining to solutions

Solutions. Definitions pertaining to solutions Slutis Defiitis pertaiig t slutis Slute is the substace that is disslved. It is usually preset i the smaller amut. Slvet is the substace that des the disslvig. It is usually preset i the larger amut. Slubility

More information

Markov processes and the Kolmogorov equations

Markov processes and the Kolmogorov equations Chapter 6 Markv prcesses ad the Klmgrv equatis 6. Stchastic Differetial Equatis Csider the stchastic differetial equati: dx(t) =a(t X(t)) dt + (t X(t)) db(t): (SDE) Here a(t x) ad (t x) are give fuctis,

More information

Claude Elysée Lobry Université de Nice, Faculté des Sciences, parc Valrose, NICE, France.

Claude Elysée Lobry Université de Nice, Faculté des Sciences, parc Valrose, NICE, France. CHAOS AND CELLULAR AUTOMATA Claude Elysée Lbry Uiversité de Nice, Faculté des Scieces, parc Valrse, 06000 NICE, Frace. Keywrds: Chas, bifurcati, cellularautmata, cmputersimulatis, dyamical system, ifectius

More information

Grade 3 Mathematics Course Syllabus Prince George s County Public Schools

Grade 3 Mathematics Course Syllabus Prince George s County Public Schools Ctet Grade 3 Mathematics Curse Syllabus Price Gerge s Cuty Public Schls Prerequisites: Ne Curse Descripti: I Grade 3, istructial time shuld fcus fur critical areas: (1) develpig uderstadig f multiplicati

More information

Active redundancy allocation in systems. R. Romera; J. Valdés; R. Zequeira*

Active redundancy allocation in systems. R. Romera; J. Valdés; R. Zequeira* Wrkig Paper -6 (3) Statistics ad Ecmetrics Series March Departamet de Estadística y Ecmetría Uiversidad Carls III de Madrid Calle Madrid, 6 893 Getafe (Spai) Fax (34) 9 64-98-49 Active redudacy allcati

More information

Lecture 21: Signal Subspaces and Sparsity

Lecture 21: Signal Subspaces and Sparsity ECE 830 Fall 00 Statistical Sigal Prcessig istructr: R. Nwak Lecture : Sigal Subspaces ad Sparsity Sigal Subspaces ad Sparsity Recall the classical liear sigal mdel: X = H + w, w N(0, where S = H, is a

More information

Directional Duality Theory

Directional Duality Theory Suther Illiis Uiversity Carbdale OpeSIUC Discussi Papers Departmet f Ecmics 2004 Directial Duality Thery Daiel Primt Suther Illiis Uiversity Carbdale Rlf Fare Oreg State Uiversity Fllw this ad additial

More information

Axial Temperature Distribution in W-Tailored Optical Fibers

Axial Temperature Distribution in W-Tailored Optical Fibers Axial Temperature Distributi i W-Tailred Optical ibers Mhamed I. Shehata (m.ismail34@yah.cm), Mustafa H. Aly(drmsaly@gmail.cm) OSA Member, ad M. B. Saleh (Basheer@aast.edu) Arab Academy fr Sciece, Techlgy

More information

Unifying the Derivations for. the Akaike and Corrected Akaike. Information Criteria. from Statistics & Probability Letters,

Unifying the Derivations for. the Akaike and Corrected Akaike. Information Criteria. from Statistics & Probability Letters, Uifyig the Derivatis fr the Akaike ad Crrected Akaike Ifrmati Criteria frm Statistics & Prbability Letters, Vlume 33, 1997, pages 201{208. by Jseph E. Cavaaugh Departmet f Statistics, Uiversity f Missuri,

More information

Design and Implementation of Cosine Transforms Employing a CORDIC Processor

Design and Implementation of Cosine Transforms Employing a CORDIC Processor C16 1 Desig ad Implemetati f Csie Trasfrms Emplyig a CORDIC Prcessr Sharaf El-Di El-Nahas, Ammar Mttie Al Hsaiy, Magdy M. Saeb Arab Academy fr Sciece ad Techlgy, Schl f Egieerig, Alexadria, EGYPT ABSTRACT

More information

Partial-Sum Queries in OLAP Data Cubes Using Covering Codes

Partial-Sum Queries in OLAP Data Cubes Using Covering Codes 326 IEEE TRANSACTIONS ON COMPUTERS, VOL. 47, NO. 2, DECEMBER 998 Partial-Sum Queries i OLAP Data Cubes Usig Cverig Cdes Chig-Tie H, Member, IEEE, Jehshua Bruck, Seir Member, IEEE, ad Rakesh Agrawal, Seir

More information

An S-type upper bound for the largest singular value of nonnegative rectangular tensors

An S-type upper bound for the largest singular value of nonnegative rectangular tensors Ope Mat. 06 4 95 933 Ope Matematics Ope Access Researc Article Jiaxig Za* ad Caili Sag A S-type upper bud r te largest sigular value egative rectagular tesrs DOI 0.55/mat-06-0085 Received August 3, 06

More information

Copyright 1978, by the author(s). All rights reserved.

Copyright 1978, by the author(s). All rights reserved. Cpyright 1978, by the authr(s). All rights reserved. Permissi t make digital r hard cpies f all r part f this wrk fr persal r classrm use is grated withut fee prvided that cpies are t made r distributed

More information

Frequency-Domain Study of Lock Range of Injection-Locked Non- Harmonic Oscillators

Frequency-Domain Study of Lock Range of Injection-Locked Non- Harmonic Oscillators 0 teratial Cferece mage Visi ad Cmputig CVC 0 PCST vl. 50 0 0 ACST Press Sigapre DO: 0.776/PCST.0.V50.6 Frequecy-Dmai Study f Lck Rage f jecti-lcked N- armic Oscillatrs Yushi Zhu ad Fei Yua Departmet f

More information

Recovery of Third Order Tensors via Convex Optimization

Recovery of Third Order Tensors via Convex Optimization Recvery f Third Order Tesrs via Cvex Optimizati Hlger Rauhut RWTH Aache Uiversity Lehrstuhl C für Mathematik (Aalysis) Ptdriesch 10 5056 Aache Germay Email: rauhut@mathcrwth-aachede Željka Stjaac RWTH

More information

ON FREE RING EXTENSIONS OF DEGREE N

ON FREE RING EXTENSIONS OF DEGREE N I terat. J. Math. & Mah. Sci. Vl. 4 N. 4 (1981) 703-709 703 ON FREE RING EXTENSIONS OF DEGREE N GEORGE SZETO Mathematics Departmet Bradley Uiversity Peria, Illiis 61625 U.S.A. (Received Jue 25, 1980) ABSTRACT.

More information

HIGH-DIMENSIONAL data are common in many scientific

HIGH-DIMENSIONAL data are common in many scientific IEEE RANSACIONS ON KNOWLEDGE AND DAA ENGINEERING, VOL. 20, NO. 10, OCOBER 2008 1311 Kerel Ucrrelated ad Regularized Discrimiat Aalysis: A heretical ad Cmputatial Study Shuiwag Ji ad Jiepig Ye, Member,

More information

Quantum Mechanics for Scientists and Engineers. David Miller

Quantum Mechanics for Scientists and Engineers. David Miller Quatum Mechaics fr Scietists ad Egieers David Miller Time-depedet perturbati thery Time-depedet perturbati thery Time-depedet perturbati basics Time-depedet perturbati thery Fr time-depedet prblems csider

More information

, the random variable. and a sample size over the y-values 0:1:10.

, the random variable. and a sample size over the y-values 0:1:10. Lecture 3 (4//9) 000 HW PROBLEM 3(5pts) The estimatr i (c) f PROBLEM, p 000, where { } ~ iid bimial(,, is 000 e f the mst ppular statistics It is the estimatr f the ppulati prprti I PROBLEM we used simulatis

More information

The Excel FFT Function v1.1 P. T. Debevec February 12, The discrete Fourier transform may be used to identify periodic structures in time ht.

The Excel FFT Function v1.1 P. T. Debevec February 12, The discrete Fourier transform may be used to identify periodic structures in time ht. The Excel FFT Fucti v P T Debevec February 2, 26 The discrete Furier trasfrm may be used t idetify peridic structures i time ht series data Suppse that a physical prcess is represeted by the fucti f time,

More information

MATH Midterm Examination Victor Matveev October 26, 2016

MATH Midterm Examination Victor Matveev October 26, 2016 MATH 33- Midterm Examiati Victr Matveev Octber 6, 6. (5pts, mi) Suppse f(x) equals si x the iterval < x < (=), ad is a eve peridic extesi f this fucti t the rest f the real lie. Fid the csie series fr

More information

Function representation of a noncommutative uniform algebra

Function representation of a noncommutative uniform algebra Fucti represetati f a cmmutative uifrm algebra Krzysztf Jarsz Abstract. We cstruct a Gelfad type represetati f a real cmmutative Baach algebra A satisfyig f 2 = kfk 2, fr all f 2 A:. Itrducti A uifrm algebra

More information

Probabilistic linguistic TODIM approach for multiple attribute decision-making

Probabilistic linguistic TODIM approach for multiple attribute decision-making Graul. Cmput. (07) : 4 DOI 0.007/s4066-07-0047-4 ORIGINAL PAPER Prbabilistic liguistic TODIM apprach fr multiple attribute decisi-makig Peide Liu Xili Yu Received: 9 April 07 / Accepted: 5 July 07 / Published

More information

MODIFIED LEAKY DELAYED LMS ALGORITHM FOR IMPERFECT ESTIMATE SYSTEM DELAY

MODIFIED LEAKY DELAYED LMS ALGORITHM FOR IMPERFECT ESTIMATE SYSTEM DELAY 5th Eurpea Sigal Prcessig Cferece (EUSIPCO 7), Pza, Plad, September 3-7, 7, cpyright by EURASIP MOIFIE LEAKY ELAYE LMS ALGORIHM FOR IMPERFEC ESIMAE SYSEM ELAY Jua R. V. López, Orlad J. bias, ad Rui Seara

More information

Information Sciences

Information Sciences Ifrmati Scieces 292 (2015) 15 26 Ctets lists available at ScieceDirect Ifrmati Scieces jural hmepage: www.elsevier.cm/lcate/is Kerel sparse represetati fr time series classificati Zhihua Che a, Wagmeg

More information

Full algebra of generalized functions and non-standard asymptotic analysis

Full algebra of generalized functions and non-standard asymptotic analysis Full algebra f geeralized fuctis ad -stadard asympttic aalysis Tdr D. Tdrv Has Veraeve Abstract We cstruct a algebra f geeralized fuctis edwed with a caical embeddig f the space f Schwartz distributis.

More information

The generation of successive approximation methods for Markov decision processes by using stopping times

The generation of successive approximation methods for Markov decision processes by using stopping times The geerati f successive apprximati methds fr Markv decisi prcesses by usig stppig times Citati fr published versi (APA): va Nue, J. A. E. E., & Wessels, J. (1976). The geerati f successive apprximati

More information

A Hartree-Fock Calculation of the Water Molecule

A Hartree-Fock Calculation of the Water Molecule Chemistry 460 Fall 2017 Dr. Jea M. Stadard Nvember 29, 2017 A Hartree-Fck Calculati f the Water Mlecule Itrducti A example Hartree-Fck calculati f the water mlecule will be preseted. I this case, the water

More information

Comparative analysis of bayesian control chart estimation and conventional multivariate control chart

Comparative analysis of bayesian control chart estimation and conventional multivariate control chart America Jural f Theretical ad Applied Statistics 3; ( : 7- ublished lie Jauary, 3 (http://www.sciecepublishiggrup.cm//atas di:.648/.atas.3. Cmparative aalysis f bayesia ctrl chart estimati ad cvetial multivariate

More information

Sound Absorption Characteristics of Membrane- Based Sound Absorbers

Sound Absorption Characteristics of Membrane- Based Sound Absorbers Purdue e-pubs Publicatis f the Ray W. Schl f Mechaical Egieerig 8-28-2003 Sud Absrpti Characteristics f Membrae- Based Sud Absrbers J Stuart Blt, blt@purdue.edu Jih Sg Fllw this ad additial wrks at: http://dcs.lib.purdue.edu/herrick

More information

5.80 Small-Molecule Spectroscopy and Dynamics

5.80 Small-Molecule Spectroscopy and Dynamics MIT OpeCurseWare http://cw.mit.edu 5.8 Small-Mlecule Spectrscpy ad Dyamics Fall 8 Fr ifrmati abut citig these materials r ur Terms f Use, visit: http://cw.mit.edu/terms. 5.8 Lecture #33 Fall, 8 Page f

More information

Preliminary Test Single Stage Shrinkage Estimator for the Scale Parameter of Gamma Distribution

Preliminary Test Single Stage Shrinkage Estimator for the Scale Parameter of Gamma Distribution America Jural f Mathematics ad Statistics, (3): 3-3 DOI:.593/j.ajms.3. Prelimiary Test Sigle Stage Shrikage Estimatr fr the Scale Parameter f Gamma Distributi Abbas Najim Salma,*, Aseel Hussei Ali, Mua

More information

Study in Cylindrical Coordinates of the Heat Transfer Through a Tow Material-Thermal Impedance

Study in Cylindrical Coordinates of the Heat Transfer Through a Tow Material-Thermal Impedance Research ural f Applied Scieces, Egieerig ad echlgy (): 9-63, 3 ISSN: 4-749; e-issn: 4-7467 Maxwell Scietific Orgaiati, 3 Submitted: uly 4, Accepted: September 8, Published: May, 3 Study i Cylidrical Crdiates

More information

Aligning Anatomy Ontologies in the Ontology Alignment Evaluation Initiative

Aligning Anatomy Ontologies in the Ontology Alignment Evaluation Initiative Aligig Aatmy Otlgies i the Otlgy Aligmet Evaluati Iitiative Patrick Lambrix, Qiag Liu, He Ta Departmet f Cmputer ad Ifrmati Sciece Liköpigs uiversitet 581 83 Liköpig, Swede Abstract I recet years may tlgies

More information

The generalized marginal rate of substitution

The generalized marginal rate of substitution Jural f Mathematical Ecmics 31 1999 553 560 The geeralized margial rate f substituti M Besada, C Vazuez ) Facultade de Ecmicas, UiÕersidade de Vig, Aptd 874, 3600 Vig, Spai Received 31 May 1995; accepted

More information

The Complexity of Translation Membership for Macro Tree Transducers

The Complexity of Translation Membership for Macro Tree Transducers The Cmplexity f Traslati Membership fr Macr Tree Trasducers Kazuhir Iaba The Uiversity f Tky kiaba@is.s.u-tky.ac.jp Sebastia Maeth NICTA ad Uiversity f New Suth Wales sebastia.maeth@icta.cm.au ABSTRACT

More information

MATHEMATICS 9740/01 Paper 1 14 Sep hours

MATHEMATICS 9740/01 Paper 1 14 Sep hours Cadidate Name: Class: JC PRELIMINARY EXAM Higher MATHEMATICS 9740/0 Paper 4 Sep 06 3 hurs Additial Materials: Cver page Aswer papers List f Frmulae (MF5) READ THESE INSTRUCTIONS FIRST Write yur full ame

More information

Matching a Distribution by Matching Quantiles Estimation

Matching a Distribution by Matching Quantiles Estimation Jural f the America Statistical Assciati ISSN: 0162-1459 (Prit) 1537-274X (Olie) Jural hmepage: http://www.tadflie.cm/li/uasa20 Matchig a Distributi by Matchig Quatiles Estimati Niklas Sgurpuls, Qiwei

More information

ON THE M 3 M 1 QUESTION

ON THE M 3 M 1 QUESTION Vlume 5, 1980 Pages 77 104 http://tplgy.aubur.edu/tp/ ON THE M 3 M 1 QUESTION by Gary Gruehage Tplgy Prceedigs Web: http://tplgy.aubur.edu/tp/ Mail: Tplgy Prceedigs Departmet f Mathematics & Statistics

More information

6.867 Machine learning, lecture 14 (Jaakkola)

6.867 Machine learning, lecture 14 (Jaakkola) 6.867 Machie learig, lecture 14 (Jaakkla) 1 Lecture tpics: argi ad geeralizati liear classifiers esebles iture dels Margi ad geeralizati: liear classifiers As we icrease the uber f data pits, ay set f

More information

Portfolio Performance Evaluation in a Modified Mean-Variance-Skewness Framework with Negative Data

Portfolio Performance Evaluation in a Modified Mean-Variance-Skewness Framework with Negative Data Available lie at http://idea.srbiau.ac.ir It. J. Data Evelpmet Aalysis (ISSN 345-458X) Vl., N.3, Year 04 Article ID IJDEA-003,3 pages Research Article Iteratial Jural f Data Evelpmet Aalysis Sciece ad

More information

Super-efficiency Models, Part II

Super-efficiency Models, Part II Super-efficiec Mdels, Part II Emilia Niskae The 4th f Nvember S steemiaalsi Ctets. Etesis t Variable Returs-t-Scale (0.4) S steemiaalsi Radial Super-efficiec Case Prblems with Radial Super-efficiec Case

More information

AP Statistics Notes Unit Eight: Introduction to Inference

AP Statistics Notes Unit Eight: Introduction to Inference AP Statistics Ntes Uit Eight: Itrducti t Iferece Syllabus Objectives: 4.1 The studet will estimate ppulati parameters ad margis f errrs fr meas. 4.2 The studet will discuss the prperties f pit estimatrs,

More information

Gusztav Morvai. Hungarian Academy of Sciences Goldmann Gyorgy ter 3, April 22, 1998

Gusztav Morvai. Hungarian Academy of Sciences Goldmann Gyorgy ter 3, April 22, 1998 A simple radmized algrithm fr csistet sequetial predicti f ergdic time series Laszl Gyr Departmet f Cmputer Sciece ad Ifrmati Thery Techical Uiversity f Budapest 5 Stczek u., Budapest, Hugary gyrfi@if.bme.hu

More information

On the affine nonlinearity in circuit theory

On the affine nonlinearity in circuit theory O the affie liearity i circuit thery Emauel Gluski The Kieret Cllege the Sea f Galilee; ad Ort Braude Cllege (Carmiel), Israel. gluski@ee.bgu.ac.il; http://www.ee.bgu.ac.il/~gluski/ E. Gluski, O the affie

More information

Control Systems. Controllability and Observability (Chapter 6)

Control Systems. Controllability and Observability (Chapter 6) 6.53 trl Systems trllaility ad Oservaility (hapter 6) Geeral Framewrk i State-Spae pprah Give a LTI system: x x u; y x (*) The system might e ustale r des t meet the required perfrmae spe. Hw a we imprve

More information

Review of Important Concepts

Review of Important Concepts Appedix 1 Review f Imprtat Ccepts I 1 AI.I Liear ad Matrix Algebra Imprtat results frm liear ad matrix algebra thery are reviewed i this secti. I the discussis t fllw it is assumed that the reader already

More information

are specified , are linearly independent Otherwise, they are linearly dependent, and one is expressed by a linear combination of the others

are specified , are linearly independent Otherwise, they are linearly dependent, and one is expressed by a linear combination of the others Chater 3. Higher Order Liear ODEs Kreyszig by YHLee;4; 3-3. Hmgeeus Liear ODEs The stadard frm f the th rder liear ODE ( ) ( ) = : hmgeeus if r( ) = y y y y r Hmgeeus Liear ODE: Suersiti Pricile, Geeral

More information

Efficient Processing of Continuous Reverse k Nearest Neighbor on Moving Objects in Road Networks

Efficient Processing of Continuous Reverse k Nearest Neighbor on Moving Objects in Road Networks Iteratial Jural f Ge-Ifrmati Article Efficiet Prcessig f Ctiuus Reverse k Nearest Neighbr Mvig Objects i Rad Netwrks Muhammad Attique, Hyug-Ju Ch, Rize Ji ad Tae-Su Chug, * Departmet f Cmputer Egieerig,

More information

Christensen, Mads Græsbøll; Vera-Candeas, Pedro; Somasundaram, Samuel D.; Jakobsson, Andreas

Christensen, Mads Græsbøll; Vera-Candeas, Pedro; Somasundaram, Samuel D.; Jakobsson, Andreas Dwladed frm vb.aau.dk : April 12, 2019 Aalbrg Uiversitet Rbust Subspace-based Fudametal Frequecy Estimati Christese, Mads Græsbøll; Vera-Cadeas, Pedr; Smasudaram, Samuel D.; Jakbss, Adreas Published i:

More information

Review for cumulative test

Review for cumulative test Hrs Math 3 review prblems Jauary, 01 cumulative: Chapters 1- page 1 Review fr cumulative test O Mday, Jauary 7, Hrs Math 3 will have a curse-wide cumulative test cverig Chapters 1-. Yu ca expect the test

More information

Physical Chemistry Laboratory I CHEM 445 Experiment 2 Partial Molar Volume (Revised, 01/13/03)

Physical Chemistry Laboratory I CHEM 445 Experiment 2 Partial Molar Volume (Revised, 01/13/03) Physical Chemistry Labratry I CHEM 445 Experimet Partial Mlar lume (Revised, 0/3/03) lume is, t a gd apprximati, a additive prperty. Certaily this apprximati is used i preparig slutis whse ccetratis are

More information

ALE 26. Equilibria for Cell Reactions. What happens to the cell potential as the reaction proceeds over time?

ALE 26. Equilibria for Cell Reactions. What happens to the cell potential as the reaction proceeds over time? Name Chem 163 Secti: Team Number: AL 26. quilibria fr Cell Reactis (Referece: 21.4 Silberberg 5 th editi) What happes t the ptetial as the reacti prceeds ver time? The Mdel: Basis fr the Nerst quati Previusly,

More information

Wavelet Video with Unequal Error Protection Codes in W-CDMA System and Fading Channels

Wavelet Video with Unequal Error Protection Codes in W-CDMA System and Fading Channels Wavelet Vide with Uequal Errr Prtecti Cdes i W-CDMA System ad Fadig Chaels MINH HUNG LE ad RANJITH LIYANA-PATHIRANA Schl f Egieerig ad Idustrial Desig Cllege f Sciece, Techlgy ad Evirmet Uiversity f Wester

More information

[1 & α(t & T 1. ' ρ 1

[1 & α(t & T 1. ' ρ 1 NAME 89.304 - IGNEOUS & METAMORPHIC PETROLOGY DENSITY & VISCOSITY OF MAGMAS I. Desity The desity (mass/vlume) f a magma is a imprtat parameter which plays a rle i a umber f aspects f magma behavir ad evluti.

More information

An Investigation of Stratified Jackknife Estimators Using Simulated Establishment Data Under an Unequal Probability Sample Design

An Investigation of Stratified Jackknife Estimators Using Simulated Establishment Data Under an Unequal Probability Sample Design Secti Survey Research Methds SM 9 A Ivestigati f Stratified ackkife Estimatrs Usig Simulated Establishmet Data Uder a Uequal Prbability Sample Desig Abstract Plip Steel, Victria McNerey, h Slata Csiderig

More information

Energy xxx (2011) 1e10. Contents lists available at ScienceDirect. Energy. journal homepage:

Energy xxx (2011) 1e10. Contents lists available at ScienceDirect. Energy. journal homepage: Eergy xxx (2011) 1e10 Ctets lists available at ScieceDirect Eergy jural hmepage: www.elsevier.cm/lcate/eergy Multi-bjective ptimizati f HVAC system with a evlutiary cmputati algrithm Adrew Kusiak *, Fa

More information

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 12

Machine Learning Theory Tübingen University, WS 2016/2017 Lecture 12 Machie Learig Theory Tübige Uiversity, WS 06/07 Lecture Tolstikhi Ilya Abstract I this lecture we derive risk bouds for kerel methods. We will start by showig that Soft Margi kerel SVM correspods to miimizig

More information

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and This article appeared i a jural published by Elsevier The attached cpy is furished t the authr fr iteral -cmmercial research ad educati use, icludig fr istructi at the authrs istituti ad sharig with clleagues

More information

Learning Similarity Measures in Non-orthogonal Space*

Learning Similarity Measures in Non-orthogonal Space* Learig Similarity Measures i N-rthgal Space* Nig Liu, Beyu Zhag, Ju Ya 3, Qiag Yag 4, Shuicheg Ya, Zheg Che, Fegsha Bai, Wei-Yig Ma Departmet Mathematical Sciece, sighua Uiversity, Beiig, 00084, PR Chia

More information

Distributed Trajectory Generation for Cooperative Multi-Arm Robots via Virtual Force Interactions

Distributed Trajectory Generation for Cooperative Multi-Arm Robots via Virtual Force Interactions 862 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS PART B: CYBERNETICS, VOL. 27, NO. 5, OCTOBER 1997 Distributed Trajectry Geerati fr Cperative Multi-Arm Rbts via Virtual Frce Iteractis Tshi Tsuji,

More information

Tactics-Based Remote Execution

Tactics-Based Remote Execution Tactics-Based Remte Executi Raesh Krisha Bala Caregie Mell Uiversity raesh@cs.cmu.edu 1 Itrducti Remte executi ca trasfrm the puiest mbile device it a cmputig giat. This wuld eable resurceitesive applicatis

More information

Bayesian Estimation for Continuous-Time Sparse Stochastic Processes

Bayesian Estimation for Continuous-Time Sparse Stochastic Processes Bayesia Estimati fr Ctiuus-Time Sparse Stchastic Prcesses Arash Amii, Ulugbek S Kamilv, Studet, IEEE, Emrah Bsta, Studet, IEEE, Michael User, Fellw, IEEE Abstract We csider ctiuus-time sparse stchastic

More information

Thermodynamic study of CdCl 2 in 2-propanol (5 mass %) + water mixture using potentiometry

Thermodynamic study of CdCl 2 in 2-propanol (5 mass %) + water mixture using potentiometry Thermdyamic study f CdCl 2 i 2-prpal (5 mass %) + water mixture usig ptetimetry Reat Tmaš, Ađelka Vrdljak UDC: 544.632.4 Uiversity f Split, Faculty f Chemistry ad Techlgy, Teslia 10/V, HR-21000 Split,

More information

TECHNICAL REPORT NO Generalization and Regularization in Nonlinear Learning Systems 1

TECHNICAL REPORT NO Generalization and Regularization in Nonlinear Learning Systems 1 DEPARTMENT OF STATISTICS Uiversity f Wiscsi 1210 West Dayt St. Madis, WI 53706 TECHNICAL REPORT NO. 1015 February 28, 2000 i Nliear Learig Systems 1 by Grace 1 Prepared fr the Hadbk f Brai Thery ad Neural

More information

Efficient Static Analysis of XML Paths and Types

Efficient Static Analysis of XML Paths and Types Efficiet Static Aalysis f XML Paths ad Types Pierre Geevès, Nabil Layaïda, Ala Schmitt T cite this versi: Pierre Geevès, Nabil Layaïda, Ala Schmitt Efficiet Static Aalysis f XML Paths ad Types Prceedigs

More information

On natural cubic splines, with an application to numerical integration formulae Schurer, F.

On natural cubic splines, with an application to numerical integration formulae Schurer, F. O atural cubic splies, with a applicati t umerical itegrati frmulae Schurer, F. Published: 0/0/970 Dcumet Versi Publisher s PDF, als kw as Versi f Recrd (icludes fial page, issue ad vlume umbers) Please

More information

Declarative approach to cyclic steady state space refinement: periodic process scheduling

Declarative approach to cyclic steady state space refinement: periodic process scheduling It J Adv Mauf Techl DOI 10.1007/s00170-013-4760-0 ORIGINAL ARTICLE Declarative apprach t cyclic steady state space refiemet: peridic prcess schedulig Grzegrz Bcewicz Zbigiew A. Baaszak Received: 16 April

More information

Literature Review of Spatio-Temporal Database Models

Literature Review of Spatio-Temporal Database Models Literature Review f Spati-Tempral Database Mdels Niks Pelekis 1,2,, Babis Thedulidis 1, Iais Kpaakis 1, Yais Thedridis 2 1 Ceter f Research i Ifrmati Maagemet (CRIM) Departmet f Cmputati, UMIST URL: http://www.crim.rg.uk

More information

COMP 551 Applied Machine Learning Lecture 11: Support Vector Machines

COMP 551 Applied Machine Learning Lecture 11: Support Vector Machines COMP 551 Applied Machine Learning Lecture 11: Supprt Vectr Machines Instructr: (jpineau@cs.mcgill.ca) Class web page: www.cs.mcgill.ca/~jpineau/cmp551 Unless therwise nted, all material psted fr this curse

More information

x 2 x 3 x b 0, then a, b, c log x 1 log z log x log y 1 logb log a dy 4. dx As tangent is perpendicular to the x axis, slope

x 2 x 3 x b 0, then a, b, c log x 1 log z log x log y 1 logb log a dy 4. dx As tangent is perpendicular to the x axis, slope The agle betwee the tagets draw t the parabla y = frm the pit (-,) 5 9 6 Here give pit lies the directri, hece the agle betwee the tagets frm that pit right agle Ratig :EASY The umber f values f c such

More information

APPLICATION OF FEM ANALYSIS METHODS TO A CYLINDER-CYLINDER INTERSECTION STRUCTURE

APPLICATION OF FEM ANALYSIS METHODS TO A CYLINDER-CYLINDER INTERSECTION STRUCTURE 18th Iteratial Cferece Structural Mechaics i Reactr echlgy (SMiR 18) Beijig, Chia, August 7-12, 25 SMiR18-F7-4 APPLICAION OF FEM ANALYSIS MEHODS O A CYLINDER-CYLINDER INERSECION SRUCURE Lipig XUE G.E.O.

More information

18.657: Mathematics of Machine Learning

18.657: Mathematics of Machine Learning 8.657: Mathematics of Machie Learig Lecturer: Philippe Rigollet Lecture 0 Scribe: Ade Forrow Oct. 3, 05 Recall the followig defiitios from last time: Defiitio: A fuctio K : X X R is called a positive symmetric

More information

An epsilon-based measure of efficiency in DEA revisited -A third pole of technical efficiency-

An epsilon-based measure of efficiency in DEA revisited -A third pole of technical efficiency- GRIPS Plicy Ifrmati Ceter Discussi Paper : 09-2 A epsil-based measure f efficiecy i DEA revisited -A third ple f techical efficiecy- Karu Te Natial Graduate Istitute fr Plicy Studies 7-22- Rppgi, Miat-ku,

More information

10-701/ Machine Learning Mid-term Exam Solution

10-701/ Machine Learning Mid-term Exam Solution 0-70/5-78 Machie Learig Mid-term Exam Solutio Your Name: Your Adrew ID: True or False (Give oe setece explaatio) (20%). (F) For a cotiuous radom variable x ad its probability distributio fuctio p(x), it

More information

The Acoustical Physics of a Standing Wave Tube

The Acoustical Physics of a Standing Wave Tube UIUC Physics 93POM/Physics 406POM The Physics f Music/Physics f Musical Istrumets The Acustical Physics f a Stadig Wave Tube A typical cylidrical-shaped stadig wave tube (SWT) {aa impedace tube} f legth

More information

WEST VIRGINIA UNIVERSITY

WEST VIRGINIA UNIVERSITY WEST VIRGINIA UNIVERSITY PLASMA PHYSICS GROUP INTERNAL REPORT PL - 045 Mea Optical epth ad Optical Escape Factr fr Helium Trasitis i Helic Plasmas R.F. Bivi Nvember 000 Revised March 00 TABLE OF CONTENT.0

More information

Dataflow Analysis and Abstract Interpretation

Dataflow Analysis and Abstract Interpretation Dataflw Analysis and Abstract Interpretatin Cmputer Science and Artificial Intelligence Labratry MIT Nvember 9, 2015 Recap Last time we develped frm first principles an algrithm t derive invariants. Key

More information

Support-Vector Machines

Support-Vector Machines Supprt-Vectr Machines Intrductin Supprt vectr machine is a linear machine with sme very nice prperties. Haykin chapter 6. See Alpaydin chapter 13 fr similar cntent. Nte: Part f this lecture drew material

More information

ESWW-2. Israeli semi-underground great plastic scintillation multidirectional muon telescope (ISRAMUTE) for space weather monitoring and forecasting

ESWW-2. Israeli semi-underground great plastic scintillation multidirectional muon telescope (ISRAMUTE) for space weather monitoring and forecasting ESWW-2 Israeli semi-udergrud great plastic scitillati multidirectial mu telescpe (ISRAMUTE) fr space weather mitrig ad frecastig L.I. Drma a,b, L.A. Pustil'ik a, A. Sterlieb a, I.G. Zukerma a (a) Israel

More information