Blind Signal Separation Methods for Integration of Neural Networks Results
|
|
- Beryl Gibson
- 6 years ago
- Views:
Transcription
1 Bnd Sgna Searaton Methods for Integraton of eura etwors Resuts Ryszard Szuu Warsaw Schoo of Economcs A.eodegosc 62, Warsaw PL Posa eefona Cyfrowa Ltd A. Jerozomse 8, Warsaw, PL Potr Wojewn Warsaw Schoo of Economcs A.eodegosc 62, Warsaw PL Posa eefona Cyfrowa Ltd A. Jerozomse 8, Warsaw, PL omasz Zabows Posa eefona Cyfrowa Ltd A. Jerozomse 8, Warsaw, PL Abstract - In ths aer t s roosed to ay bnd sgna searaton methods to mrove a neura networ redcton. Resuts generated by any regresson mode usuay ncude both constructve and destructve comonents. In case of a few modes, some of the comonents can be common to a of them. Our am s to fnd the bass eements and dstngush the comonents wth the constructve nfuence on the modeng quaty from the destructve ones. After rejectng the destructve eements from the modes resuts t s observed the enhancement of the resuts n terms of some standard error crtera. he vadty and hgh erformance of the concet s resented on the rea robem of energy oad redcton. Keywords: ensembe methods, regresson, neura networs, bnd sgna searaton. Introducton he neura networs, as we as the other regresson modes, try to reresent the deendency between nut data and target. o obtan roer resut, the cost functon of the networ arameters s otmzed [4, 2]. ycay, n rea robems many dfferent tyes and structures of neura networs are tested. After the earnng stage s fnshed we can evauate redcton resuts and choose the best neura mode [,3,6,9]. In ths standard methodoogy there aear some robems that shoud be consdered. Dfferent cost functons can ead to varous networ arameters as the otma and sometmes the choce of the functon s not obvous. here s no assurance that earnng rocess reay fnds the goba otmum of the cost functon [2,4]. Fnay, the term the best neura networ s not recse, because there est many evauaton crtera and dfferent error measures can ndcate dfferent modes as the best. It s qute ossbe that there est few modes of smar quaty or accordng to dfferent crtera each of the modes s evauated as the best. herefore, t seems natura to ntegrate and to use the nformaton generated by many neura networs what eads us to wde area of ensembe methods. Usuay soutons roose the combnaton of a few modes by mng ther resuts or arameters [5,3,22]. In ths aer we roose an aternatve concet based on mutdmensona decomostons [2]. ow et s assume that we have severa neura networs of accetabe redcton quaty. If dfferent networs gve reatvey good resuts t means that a of them are ocated cose to the target and the resuts of every mode ossess the constructve comonents as we as the destructve comonents and noses. We can assume that there est the eements common to a the modes. he common constructve comonents are assocated wth the unnown true vaue, of course. On the other hand the destructve comonents can be resent due to many reasons. Some of them ncude: mssng data, not recse arameter estmaton and dstrbuton assumtons. Our am s to dentfy the destructve eements and to emnate them what shoud gve us the modeng mrovement. o obtan such nterestng comonents we coect artcuar mode resuts together, treat them as mutvarate varabe and transform them wth one of the bnd sgna searaton methods. he fu rocess has the foowng stes:. Coectng the modes resuts together n one mutvarate varabe 2. Decomoston of mutvarate varabe nto bass comonents by bnd sgna searaton methods 3. Identfcaton and emnaton of destructve comonents from bass sgnas 4. Returnng from ceaned bass sgnas to ceaned redcton resuts, n smest case by transformaton nverse to decomoston 5. Choosng the best resuts from ceaned redcton resuts he above methodoogy s resented n neura networ redcton framewor, but t can be addressed to any other regresson modes. 2 Mode resuts ntegraton We assume that after earnng rocess each neura networ resut ncudes two tyes of comonents: constructve assocated wth the target and destructve
2 assocated wth the naccurate earnng data, ndvdua roertes of modes etc. Many of the comonents are common to a the modes due to the same target varabe, earnng data set, smar mode structures or otmzaton methods. ow we eress mode resuts not n terms of nut data but n terms of mentoned above constructve and destructve atent comonents. Let the resut of -th snge outut neura networ, =,..., m, wth observatons, be a functon of constructve mact comonents t, t 2,..., t, and destructve comonents v, v2,..., vq, what gves = F t, t..., t, v, v..., v, ( ( 2 2, q where F s a nage functon what can be treated as mng system ether. For smcty of further anayss we assume near mng system, what gves us redcton resuts as = α... + t α t + β v + β qvq. (2 In cose matr form we have where: = [ t t,..., t ], 2 = α t + β v, (3 t s a matr of target comonents, = [ v v,..., ] v s a q matr of, 2 v q [ α,..., α ], = [ β,..., β q resduas, α = β ] are vectors of coeffcents. In case of many modes we have where = [,..., ] = As, (4, 2 m s a m matr of mode t resuts, s = s a n matr of bass comonents v α β ( n = + q, and A = M M s m n matr of α m β m mng coeffcents. Fgure : System for ntegraton of neura networ resuts A the neura networ resuts taen together gve us one mutvarate varabe whch ncudes atent comonents assocated wth dfferent roertes of artcuar resuts. Our am s to dentfy bass comonents (sgnas and mng matr from observed modes resuts, and reject the destructve comonents v (reacng them by zero to t obtan ŝ =, what gves us urfed target estmaton 0 t ˆ = Asˆ = A. 0 he effectveness of the method hghy deends on the roer dstncton t from v. hs tas requres some decson system, see Fg.. he smest souton s to chec the mact of each bass comonent and ther combnatons on the fna resuts by rejectng them one by one and mng the rest n transformaton nverse to decomoston system. he one of cruca onts n our methodoogy s the roer decomoston choce. 3 Bnd Sgna Searaton and decomoston agorthms From many transformatons whch can be used to obtan bass sgnas we focus on bnd sgnas searaton (BSS. hese methods am at dentfcaton of the unnown sgnas med n the unnown system [7,4,20]. In our case, we are not oong for secfc rea sgnas but rather for nterestng anaytca reresentaton. he BSS methods use dfferent features and roertes of data but most of them can be consdered as oong for data reresentaton of the form (4. o fnd the atent varabes A and s we can use a transformaton defned n m by matr W R, such that (5 y = W, (6 where y s reated to s and satsfes the foowng reaton y = PDs, (7 where P s a ermutaton matr and D s a dagona matr [7,4,7]. he reaton (7 means that estmated sgnas can be rescaed and reordered n comarson to orgna sources. In our case t s not cruca, therefore y can be treated drecty as estmated verson of sources s. here are some dfferent addtona assumtons deendng on artcuar BSS method [7]. We focus on methods based on decorreaton and ndeendent comonent anayss. In further consderaton we assume, for smcty, that n=m. Decorreaton s one of the most ouar statstca rocedures for the emnaton of the deendences estng n the data. In ractce, decorreaton can be erformed by dagonazaton of the correaton matr { ( } R = E ( ( (. (8 = herefore we need such matr W that for transformaton (6 we have R yy = WR W = E, (9
3 where E s any dagona matr. From many methods eadng to the decorreaton matr W (abe the mortant roe ays rnca comonent anayss due to ts nterestng statstca and agebrac roertes [5]. abe : Methods of decorreaton ossbe for modes decomoston Method a Factorsaton Decorreaton Form correaton / 2 / 2 R = R R / 2 W = R LU R = LU W = L Choesy R = G G W = G EIG (PCA R = UΣ U W = U a Defnton and characterstc of the artcuar methods can be found n [0]. Standard decorreaton s very usefu anaytca too but n the contet of the BSS robem t s rather suortng method because of ts sma searaton abtes and t s usuay used as rerocessng before the man agorthms [6,7]. In our case, the most nterestng thng s a romsng anaytca reresentaton of sgnas and therefore, decorreaton can ay an mortant roe n resented methodoogy. Indeendent comonent anayss (ICA s a statstca too, whch aows decomoston of observed varabes nto ndeendent comonents [6,9,7]. yca agorthms for ICA eore hgher order statstca deendences n a dataset, so after ICA decomoston we have got sgnas (varabes wthout any near and non-near statstca deendences [2,6]. hs s the man dfference from the standard correaton methods (e.g. PCA, whch aow us to anayze ony the near deendences. o obtan ndeendent comonents we need to anayze statstca structure of y. he jont robabty of ndeendent varabes can be factorzed by the roduct of the margna robabtes q ( y ( y y y ( y ( y K ( y = ( y, y,..., y. 2 2 n n... n 2 n (0 here are many ways to obtan (0. One of the most ouar methods s based on measure of dfference between y (y and q y (y by the Kubac-Leber dvergence [2,7] D KL y ( y ( y ( y qy ( y = y ( yog dy. ( q ( y + We are oong for a matr W that mnmzes ( so W ot = mn D KL ( y ( W qy ( W. (2 W here are many agorthms for ICA. wo of them are atura Gradent Agorthm [ I E{ f ( y( y ( }] W( W( = µ ( and Fed Pont Agorthm [ E{ f ( y( y ( } dag( E{ f ( y y }] W( y, (3 W( = D.(4 ' Where D = dag( / E{ f ( y y } E{ f ( y }, and f ( y = [ f ( y,..., f ] n ( y n s a vector of nonneartes wth otma form of og( ( y f ( y =. (5 y Otma nonneartes used n (5 need the nowedge of source robabty dstrbutons what s mossbe n genera. here are many roosas to sove ths robem e arametrc and nonarametrc densty estmaton, adatve earnng choce. For unmoda dstrbutons the anayss of urtoss arameters gves us a sme heurstc formua for nonneartes choce [8] tanh( β y r f ( y = sgn( y y y κ 4( y > 0 κ 4( y < 0, (6 κ 4( y = 0 where r >, β are constants and normazed urtoss κ s gven by = κ ( y E{ y }/ E { y } 3. (7 Smooth Comonent Anayss (SmCA s a method of fndng the smooth comonents n a mutvarate varabe wth temora structure [7]. here are varous smoothness measures for BSS e e.g. the Stone s redctabty measure [23], but t s assocated wth the arbtrary choce of movng average form. We roose a new smoothness measure as P ( y = y ( y (, R + δ ( R = 2 (8 where R = ma( y mn( y means range and symbo δ (. means Kronecer deta. he measure P ( y has sme nterretaton: s mama when changes n each ste are equa to range (mama, and s mnma when data are constant. he ossbe vaues are from to 0. o obtan a dfferentabe smoothness measure we can aromate (8 by P ( y = 2 og(cosh( y ( y ( = 2, og(cosh( R + δ ( R (9 where the og(cosh(. functon s used as the aromaton of absoute vaue functon. Our am s to fnd such W = [ w, w 2,..., w3] that for y = W we obtan y = [ y, y2,..., yn ] where y = w mamzes w w = = arg ma( P( w, (20 Havng estmated the frst - smooth comonents the net one s cacuated as most smooth comonents of the resdua
4 = w = arg ma( P ( w ( ww, (2 w = As the numerca agorthm for fndng (2 we can tae ewton method wth mute startng ont run. From many resuts we choose the one wth mama P vaue. In many cases the SmCA gve resuts smar to ICA or other BSS methods but n contradstncton to them t s focused on temora structure of data, what s mortant for robems descrbed by tme seres. 4 Decson system and mutcrtera choce Presented method assumes estence and utzaton of some constructve comonents for each mode resut. he estmaton of the comonents requres choce of the transformaton and abeng of the resutng comonents as the destructve or constructve. hs can be dffcut tas because obtaned comonents mght be not ure constructve or destructve. More recsey, artcuar comonent can have constructve mact on one mode and destructve on the other. Moreover, there may est comonents destructve as a snge but constructve n a grou. herefore, for each transformaton we have to chec the mact of a the comonents subsets on the fna resuts. he smest method for choosng the transformaton and abe ts comonents s as foows. For each transformaton emnate artcuar comonents subsets, combne the remanng ones by mng matr and evauate the fna resuts accordng to chosen crteron. he most sutabe transformaton and abeng woud be the one that ead to the best resuts accordng to the artcuar crteron. he other method can tae advantage of usng many crtera. he mutcrtera aroach s acabe because BSS transformatons gve somethng hysca comonents and the emnaton of the rea hysca nose shoud mrove many crtera. It means utzaton of dfferent tye nformaton e varance, senstvty to outers etc. Whereas the dfferent crtera have dfferent measure characterstcs e scaes, sewness, urtoss etc., any sme aggregaton s not recommended. We roose a decson systems based on Pareto-otmaty of modes resuts n mutdmensona sace of crtera. Let us assume we have a few modes, =,...,m, a set of varous transformatons Φ, =,..., mφ, and some quaty crtera C, =,...,mc. Each transformaton Φ searates some source sgnas S (n the number of m m that can be abeed n 2 -way as constructve or m destructve V ones, S = [, V ], =,..., 2. Let ˆ ( denote the mode resuts based on constructve sgnas obtaned n transformaton Φ, and C ( ( ˆ be the vaue of error measure C for the ˆ. resut ( he choce of the best transformaton Φ and abeng w foow the rue. Let denote the number of Pareto-otma modes resutng from the transformaton Φ and the abeng, where the Pareto-otmaty deends on the crtera C, =,...,m : = card{ 0, 0 = ( ˆ ( ˆ 0 C ( C ( K m 0 C, ( ˆ ( ( ˆ 0 C < C ( = K m 0 C = Km }. C (22 he best constructve sgnas * gve the greatest * * number of Pareto-otma modes ˆ ( *,,...,m =, so: * * * = { * = ma }. (23 * If there are a few best transformatons and abengs * the fna choce s eft to the anayst. It shoud be noted that transformaton choce and comonent dentfcaton can be done on data for whch we have the target, but fna decomoston and mng matr cacuaton shoud contan test data (or redcton resuts when the target s not avaabe., 5 Generazed mng After the destructve comonents are dentfed and emnated the urfed modes resut can be obtaned by ˆ = Asˆ = A[ t 0]. he reacement of the destructve sgnas by zero s equvaent to uttng zero n corresondng to them mng coeffcents. If we eress the mng matr as A = [ a, a2,..., a n ] the urfed resuts can be descrbed as ˆ Asˆ A t 0 Aˆ = = [ ] = [ t v] = As ˆ where A ˆ [ a, a,..., a, 0, 0,..., ] = n (24 Fgure 2: Imrovement rocess based on decomoston he mng matr  s the best matr we can fnd by sme test wth emnatng each combnaton of the comonents. As t was mentoned above the comonents can be not ure, so ther mact shoud have weght other than 0. It means that we can try to fnd the better mng system than descrbed by A (nverse to decomoston system. he new mng system can be formuated even more genera than sme near. We can tae MLP neura networ as the mng system
5 = g B [ g ( B s + b ] +, (25 2( 2 b2 where g (. s a vector of nonneartes, B s a weght matr and b s a bas vector resectvey for -th ayer, =,2. he frst weght ayer w roduce resuts reated to (24 when we tae B Aˆ =. But we can eect that net weghts ayers and nonneartes can mrove resut due to earnng rocess. In other words, n the earnng rocess we search for better mng system startng from system descrbed by  wth nta weghts of B (0 Aˆ =, see Fg.2. 6 Practca eerment In ths aer we anayse the rea robem of forecastng the houry eectrcty consumton for 24 hours ahead basng on the hstorca energy oad and caendar [], [8]. he data reresent eectrcty consumton n Poand durng 4377 days n 80 s and 90 s. he seres from 987 t 996 are randomy dvded nto earnng (2008 days and vadatng set (2007 days and data of 997 (362 days as a testng set. he erformance of the redctons made at 0 a.m. each day s measured by MSE and MAPE crtera [3]. abe 2. he error vaues obtaned by modes mroved by decomoston agorthms. he rates of error reducton are n bracets MAPE[%] MSE [0-4 ] Base a,5523 4,805 ICA,5495 4,349 (0,8% (,09% ICA+,575 3,9783 (2,24% (4,84% PCA,507 3,8929 (3,26% (6,88% PCA+,53 3,9208 (2,64% (6,2% Choesy,5028 3,885 (3,9% (7,07% Choesy+,548 3,9404 (2,42% (5,74% SmCA,530 4,0000 (2,53% (4,32% SmCA+,502 3,9008 (2,7% (6,69% a Best score among the rmary mode resuts: MLP 64 More recse concuson can be drawn from the best scores of the modes before and after decomostons. As we can see the decomoston agorthms enabe mrovement of the resuts by 3% (MAPE to 7% (MSE. If the decomoston agorthm manages to etract ure comonents (PCA and Choesy, the genera mng (neura mode s not better than the sme mng. But n case of neffectve decomoston (ICA we can observe that the genera mng can mrove the resuts. 7 Concusons Fgure 3: Sames of bass sgnas etracted usng PCA. Sgnas y-y3 were dentfed as constructve and y4-y6 as destructve for the mode quaty Seven MLP neura networs wth 60, 62, 64, 66, 68, 70, and 72 neurons n second hdden ayer were traned. he bass sgnas etracton was erformed usng PCA (Fg.3, Choesy and ICA agorthms. he constructve comonents were dentfed usng decson system emoyng Pareto-otmaty condton descrbed n Secton 4. Fgure 4: Rea energy oad and PCA mroved redcton he rea sgna (target and the best mroved modes resuts are shown n Fg.4. We can observe that the redcton systems refects qute we the features of rea rocess. In ths artce we resent a new aroach to the concet of redcton mrovement, where the nformaton from few neura networs s ntegrated. We eore the fact, that there can be many neura networ modes wth dfferent roertes, gvng us dfferent nformaton about redcton robem. Varous crtera can be used for mode resuts evauaton and t s chaengng tas to fnd the sueror mode, because the crtera refect dfferent asects of the robem. On the other hand, even f we have the best mode accordng to our man crteron t can be st vauabe to utze the sghty worse modes for further redcton mrovement. We roose the methodoogy of mode resuts ntegraton based on dentfcaton and emnaton destructve comonents common to many modes. o fnd such comonents we roose BSS methods but the other methods e agebrac decomostons can be used too. Our aroach hes to mrove the modes accuracy and enhances ther generazaton abtes due to combnng the constructve comonents common to a the neura networs. he ractca eerment wth the energy oad redcton confrms vadty of our method. he assumton of modes resuts as a near combnaton of the bass comonents can be etended on nonnear or dynamc systems or even qute omtted n consderatons what eads to machne anayss of many transformatons and sgnas. In such case we have to accet that the
6 transformaton s not assocated wth the mng system of resuts and there s no straghtforward nterretaton of the source sgnas. References [] H. Aae, A new oo at the statstca mode dentfcaton, IEEE ransactons on Automatc Contro, Vo.9, 974, [2] S. Amar, A. Cchoc, and H.H. Yang, A new earnng agorthm for bnd sgna searaton, Advances n eura Informaton Processng Systems IPS-995, MI Press, Cambrdge MA, 996, [3] J.S. Armstrong, F. Cooy, Error Measures For Generazng About Forecastng Methods: Emrca Comarsons, Internatona Journa of Forecastng, Vo.8, 992, [4] C.M. Bsho, eura networs for attern recognton, Oford Unv. Press, Oford UK, 996. [5] L. Breman, Baggng redctors, Machne Learnng, Vo.24, 996, [6] J.F. Cardoso, Hgh-order contrasts for ndeendent comonent anayss, eura Comutaton, Vo., 999, [7] A. Cchoc, S. Amar, Adatve Bnd Sgna and Image Processng, John Wey, Chchester, [8] A. Cchoc, I. Sabaa, S. Cho, B. Orser, and R. Szuu, Sef adatve ndeendent comonent anayss for sub-gaussan and suer-gaussan mtures wth unnown number of sources and addtve nose, Proc.of OLA-97, Vo.2, Hawa USA, 997, [9] P. Comon, Indeendent comonent anayss, a new concet?, Sgna Processng, Vo.36, Esever, 994. [0] G.H. Goub, C.F. Van-Loan, Matr Comutatons, Johns Hons, 3rd ed., 996. [] A. Harvey, S. Kooman, Forecastng Houry Eectrcty Demand Usng me-varyng Snes, Journa of the Amercan Statstca Ass., Vo.88, 993, [2] S. Hayn, eura networs: a comrehensve foundaton, Macman, ew Yor, 994. [3] J. Hoetng, D. Madgan, A. Raftery, and C. Vonsy, Bayesan mode averagng: a tutora, Statstca Scence, Vo.4, 999, [4] A. Hyvärnen, J. Karhunen, E. Oja, Indeendent Comonent Anayss, John Wey, 200. [5]. Joffe, Prnca Comonent Anayss, Srnger- Verag, 986. [6] R.L. Kennedy (ed., Y. Lee, B. Van Roy, C. Reed, and R.P. Lman, Sovng Data Mnng Probems wth Pattern Recognton, Prentce Ha, 997. [7] -W. Lee, M. Groam, A.J. Be, and.j. Sejnows, A Unfyng Informaton-theoretc Framewor for Indeendent Comonent Anayss, Comuters & Mathematcs wth Acatons, Vo.3, 2000,. -2. [8] A. Lendasse, M. Cottre, V. Wertz., M. Verdeysen, Predcton of Eectrc Load usng Kohonen Mas Acaton to the Posh Eectrcty Consumton, Proc. Am. Contro Conf., Anchorage AK, 2002, [9]. Mtche, Machne Learnng, McGraw-H, Boston, 997. [20] R. Szuu, A. Cchoc, Bnd sgna searaton usng second order statstcs, Proc. SPEO 0, 200, [2] R. Szuu, P. Wojewn, and. Zabows, Mode Imrovement by the Statstca Decomoston, Proc. ICAISC 04, Lecture otes n Comuter Scence, Srnger-Verag, Hedeberg, 2004, [22] Y. Yang, Adatve regresson by mng, Journa of Amercan Statstca Assocaton, Vo.96, 200. [23] 5. J.V. Stone, Bnd Source Searaton Usng emora Predctabty, eura Comutaton, 3(7, Juy, 200,
REMODELLING OF VIBRATING SYSTEMS VIA FREQUENCY-DOMAIN-BASED VIRTUAL DISTORTION METHOD
REMODELLING OF VIBRATING SYSTEMS VIA FREQUENCY-DOMAIN-BASED VIRTUAL DISTORTION METHOD Małgorzata MRÓZ and Jan HOLNICKI-SZULC Insttute of Fundamenta Technoogca Research, Swetokrzyska 21, -9 Warsaw, Poand
More informationNeural network-based athletics performance prediction optimization model applied research
Avaabe onne www.jocpr.com Journa of Chemca and Pharmaceutca Research, 04, 6(6):8-5 Research Artce ISSN : 0975-784 CODEN(USA) : JCPRC5 Neura networ-based athetcs performance predcton optmzaton mode apped
More informationResearch on Complex Networks Control Based on Fuzzy Integral Sliding Theory
Advanced Scence and Technoogy Letters Vo.83 (ISA 205), pp.60-65 http://dx.do.org/0.4257/ast.205.83.2 Research on Compex etworks Contro Based on Fuzzy Integra Sdng Theory Dongsheng Yang, Bngqng L, 2, He
More informationMARKOV CHAIN AND HIDDEN MARKOV MODEL
MARKOV CHAIN AND HIDDEN MARKOV MODEL JIAN ZHANG JIANZHAN@STAT.PURDUE.EDU Markov chan and hdden Markov mode are probaby the smpest modes whch can be used to mode sequenta data,.e. data sampes whch are not
More informationThe Use of Principal Components Analysis in the Assessment of Process Capability Indices
Jont Statstca Meetngs - Secton on Physca & Engneerng Scences (SPES) The Use of Prnca omonents Anayss n the Assessment of Process aabty Indces Evdoka Xekaak Mchae Peraks Deartment of Statstcs Athens Unversty
More informationImage Classification Using EM And JE algorithms
Machne earnng project report Fa, 2 Xaojn Sh, jennfer@soe Image Cassfcaton Usng EM And JE agorthms Xaojn Sh Department of Computer Engneerng, Unversty of Caforna, Santa Cruz, CA, 9564 jennfer@soe.ucsc.edu
More informationCOXREG. Estimation (1)
COXREG Cox (972) frst suggested the modes n whch factors reated to fetme have a mutpcatve effect on the hazard functon. These modes are caed proportona hazards (PH) modes. Under the proportona hazards
More informationXin Li Department of Information Systems, College of Business, City University of Hong Kong, Hong Kong, CHINA
RESEARCH ARTICLE MOELING FIXE OS BETTING FOR FUTURE EVENT PREICTION Weyun Chen eartment of Educatona Informaton Technoogy, Facuty of Educaton, East Chna Norma Unversty, Shangha, CHINA {weyun.chen@qq.com}
More informationSupplementary Material: Learning Structured Weight Uncertainty in Bayesian Neural Networks
Shengyang Sun, Changyou Chen, Lawrence Carn Suppementary Matera: Learnng Structured Weght Uncertanty n Bayesan Neura Networks Shengyang Sun Changyou Chen Lawrence Carn Tsnghua Unversty Duke Unversty Duke
More informationExample: Suppose we want to build a classifier that recognizes WebPages of graduate students.
Exampe: Suppose we want to bud a cassfer that recognzes WebPages of graduate students. How can we fnd tranng data? We can browse the web and coect a sampe of WebPages of graduate students of varous unverstes.
More informationComparison of BPA and LMA methods for Takagi - Sugeno type MIMO Neuro-Fuzzy Network to forecast Electrical Load Time Series
Comarson of BPA and LA methods for akag - Sugeno tye IO euro-fuzzy etwork to forecast Eectrca Load me Seres [Fex Pasa] Comarson of BPA and LA methods for akag - Sugeno tye IO euro-fuzzy etwork to forecast
More informationQuality-of-Service Routing in Heterogeneous Networks with Optimal Buffer and Bandwidth Allocation
Purdue Unversty Purdue e-pubs ECE Technca Reorts Eectrca and Comuter Engneerng -6-007 Quaty-of-Servce Routng n Heterogeneous Networs wth Otma Buffer and Bandwdth Aocaton Waseem Sheh Purdue Unversty, waseem@urdue.edu
More informationDr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur
Analyss of Varance and Desgn of Exerments-I MODULE III LECTURE - 2 EXPERIMENTAL DESIGN MODELS Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 2 We consder the models
More informationAssociative Memories
Assocatve Memores We consder now modes for unsupervsed earnng probems, caed auto-assocaton probems. Assocaton s the task of mappng patterns to patterns. In an assocatve memory the stmuus of an ncompete
More informationNonextensibility of energy in Tsallis statistics and the zeroth law of
Nonextensbty of energy n Tsas statstcs and the zeroth a of thermodynamcs onge Ou and Jncan hen* T Word Laboratory, P. O. 870, eng 00080, Peoe s Reubc of hna and Deartment of Physcs, Xamen nversty, Xamen
More informationAn LSB Data Hiding Technique Using Prime Numbers
An LSB Data Hdng Technque Usng Prme Numbers Sandan Dey (), Aj Abraham (), Sugata Sanya (3) Anshn Software Prvate Lmted, Kokata 79 Centre for Quantfabe Quaty of Servce n Communcaton Systems Norwegan Unversty
More informationReconstruction History. Image Reconstruction. Radon Transform. Central Slice Theorem (I)
Reconstructon Hstory wth Bomedca Acatons EEG-475/675 Prof. Barner Reconstructon methods based on Radon s wor 97 cassc mage reconstructon from roectons aer 97 Hounsfed deveo the frst commerca x-ray CT scanner
More informationNeural-Network-Based Fuzzy Group Forecasting with Application to Foreign Exchange Rates Prediction
Neura-Network-Based Fuy Grou Forecastng wth Acaton to Foregn Echange Rates Predcton Lean Yu,, Kn Keung La, and Shouyang Wang Insttute of Systems Scence, Academy of Mathematcs and Systems Scence, Chnese
More informationController Design of Nonlinear TITO Systems with Uncertain Delays via Neural Networks and Error Entropy Minimization
Controer Desgn of Nonnear TITO Systes wth Uncertan Deays va Neura Networs Error Entroy Mnzaton J. H. Zhang A. P. Wang* H. Wang** Deartent of Autoaton North Chna Eectrc Power Unversty Bejng 6 P. R. Chna
More informationConfidence intervals for weighted polynomial calibrations
Confdence ntervals for weghted olynomal calbratons Sergey Maltsev, Amersand Ltd., Moscow, Russa; ur Kalambet, Amersand Internatonal, Inc., Beachwood, OH e-mal: kalambet@amersand-ntl.com htt://www.chromandsec.com
More informationDmitry A. Zaitsev Odessa National Telecommunication Academy Kuznechnaya, 1, Odessa, Ukraine
th Worksho on Agorthms and Toos for Petr Nets, Setember - October, 4, Unversty of Paderborn, Germany, 75-8 Sovng the fundamenta equaton of Petr net usng the decomoston nto functona subnets Dmtry A Zatsev
More informationNested case-control and case-cohort studies
Outne: Nested case-contro and case-cohort studes Ørnuf Borgan Department of Mathematcs Unversty of Oso NORBIS course Unversty of Oso 4-8 December 217 1 Radaton and breast cancer data Nested case contro
More informationIndependent Component Analysis
Indeendent Comonent Analyss Mture Data Data that are mngled from multle sources May not now how many sources May not now the mng mechansm Good Reresentaton Uncorrelated, nformaton-bearng comonents PCA
More informationPredicting Model of Traffic Volume Based on Grey-Markov
Vo. No. Modern Apped Scence Predctng Mode of Traffc Voume Based on Grey-Marov Ynpeng Zhang Zhengzhou Muncpa Engneerng Desgn & Research Insttute Zhengzhou 5005 Chna Abstract Grey-marov forecastng mode of
More informationMultispectral Remote Sensing Image Classification Algorithm Based on Rough Set Theory
Proceedngs of the 2009 IEEE Internatona Conference on Systems Man and Cybernetcs San Antono TX USA - October 2009 Mutspectra Remote Sensng Image Cassfcaton Agorthm Based on Rough Set Theory Yng Wang Xaoyun
More informationBoundary Value Problems. Lecture Objectives. Ch. 27
Boundar Vaue Probes Ch. 7 Lecture Obectves o understand the dfference between an nta vaue and boundar vaue ODE o be abe to understand when and how to app the shootng ethod and FD ethod. o understand what
More informationDr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur
Analyss of Varance and Desgn of Exerments-I MODULE II LECTURE - GENERAL LINEAR HYPOTHESIS AND ANALYSIS OF VARIANCE Dr. Shalabh Deartment of Mathematcs and Statstcs Indan Insttute of Technology Kanur 3.
More informationA General Class of Selection Procedures and Modified Murthy Estimator
ISS 684-8403 Journal of Statstcs Volume 4, 007,. 3-9 A General Class of Selecton Procedures and Modfed Murthy Estmator Abdul Bast and Muhammad Qasar Shahbaz Abstract A new selecton rocedure for unequal
More informationLINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity
LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 31 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 6. Rdge regresson The OLSE s the best lnear unbased
More informationMACHINE APPLIED MACHINE LEARNING LEARNING. Gaussian Mixture Regression
11 MACHINE APPLIED MACHINE LEARNING LEARNING MACHINE LEARNING Gaussan Mture Regresson 22 MACHINE APPLIED MACHINE LEARNING LEARNING Bref summary of last week s lecture 33 MACHINE APPLIED MACHINE LEARNING
More informationThe Application of BP Neural Network principal component analysis in the Forecasting the Road Traffic Accident
ICTCT Extra Workshop, Bejng Proceedngs The Appcaton of BP Neura Network prncpa component anayss n Forecastng Road Traffc Accdent He Mng, GuoXucheng &LuGuangmng Transportaton Coege of Souast Unversty 07
More informationAn Accurate Heave Signal Prediction Using Artificial Neural Network
Internatonal Journal of Multdsclnary and Current Research Research Artcle ISSN: 2321-3124 Avalale at: htt://jmcr.com Mohammed El-Dasty 1,2 1 Hydrograhc Surveyng Deartment, Faculty of Martme Studes, Kng
More informationNumerical integration in more dimensions part 2. Remo Minero
Numerca ntegraton n more dmensons part Remo Mnero Outne The roe of a mappng functon n mutdmensona ntegraton Gauss approach n more dmensons and quadrature rues Crtca anass of acceptabt of a gven quadrature
More informationShort-Term Load Forecasting for Electric Power Systems Using the PSO-SVR and FCM Clustering Techniques
Energes 20, 4, 73-84; do:0.3390/en40073 Artce OPEN ACCESS energes ISSN 996-073 www.mdp.com/journa/energes Short-Term Load Forecastng for Eectrc Power Systems Usng the PSO-SVR and FCM Custerng Technques
More informationQUARTERLY OF APPLIED MATHEMATICS
QUARTERLY OF APPLIED MATHEMATICS Voume XLI October 983 Number 3 DIAKOPTICS OR TEARING-A MATHEMATICAL APPROACH* By P. W. AITCHISON Unversty of Mantoba Abstract. The method of dakoptcs or tearng was ntroduced
More informationMODEL TUNING WITH THE USE OF HEURISTIC-FREE GMDH (GROUP METHOD OF DATA HANDLING) NETWORKS
MODEL TUNING WITH THE USE OF HEURISTIC-FREE (GROUP METHOD OF DATA HANDLING) NETWORKS M.C. Schrver (), E.J.H. Kerchoffs (), P.J. Water (), K.D. Saman () () Rswaterstaat Drecte Zeeand () Deft Unversty of
More informationComparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method
Appled Mathematcal Scences, Vol. 7, 0, no. 47, 07-0 HIARI Ltd, www.m-hkar.com Comparson of the Populaton Varance Estmators of -Parameter Exponental Dstrbuton Based on Multple Crtera Decson Makng Method
More informationLINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity
LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have
More informationFuzzy approach to solve multi-objective capacitated transportation problem
Internatonal Journal of Bonformatcs Research, ISSN: 0975 087, Volume, Issue, 00, -0-4 Fuzzy aroach to solve mult-objectve caactated transortaton roblem Lohgaonkar M. H. and Bajaj V. H.* * Deartment of
More informationComparing two Quantiles: the Burr Type X and Weibull Cases
IOSR Journal of Mathematcs (IOSR-JM) e-issn: 78-578, -ISSN: 39-765X. Volume, Issue 5 Ver. VII (Se. - Oct.06), PP 8-40 www.osrjournals.org Comarng two Quantles: the Burr Tye X and Webull Cases Mohammed
More informationClassification Bayesian Classifiers
lassfcaton Bayesan lassfers Jeff Howbert Introducton to Machne Learnng Wnter 2014 1 Bayesan classfcaton A robablstc framework for solvng classfcaton roblems. Used where class assgnment s not determnstc,.e.
More informationEstimation of a proportion under a certain two-stage sampling design
Etmaton of a roorton under a certan two-tage amng degn Danutė Kraavcatė nttute of athematc and nformatc Lthuana Stattc Lthuana Lthuana e-ma: raav@tmt Abtract The am of th aer to demontrate wth exame that
More informationWeb-Mining Agents Probabilistic Information Retrieval
Web-Mnng Agents Probablstc Informaton etreval Prof. Dr. alf Möller Unverstät zu Lübeck Insttut für Informatonssysteme Karsten Martny Übungen Acknowledgements Sldes taken from: Introducton to Informaton
More information2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification
E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton
More informationChapter 9: Statistical Inference and the Relationship between Two Variables
Chapter 9: Statstcal Inference and the Relatonshp between Two Varables Key Words The Regresson Model The Sample Regresson Equaton The Pearson Correlaton Coeffcent Learnng Outcomes After studyng ths chapter,
More informationDeriving the Dual. Prof. Bennett Math of Data Science 1/13/06
Dervng the Dua Prof. Bennett Math of Data Scence /3/06 Outne Ntty Grtty for SVM Revew Rdge Regresson LS-SVM=KRR Dua Dervaton Bas Issue Summary Ntty Grtty Need Dua of w, b, z w 2 2 mn st. ( x w ) = C z
More informationChapter 6. Rotations and Tensors
Vector Spaces n Physcs 8/6/5 Chapter 6. Rotatons and ensors here s a speca knd of near transformaton whch s used to transforms coordnates from one set of axes to another set of axes (wth the same orgn).
More informationNonlinear Robust Regression Using Kernel Principal Component Analysis and R-Estimators
IJCSI Internatona Journa of Comuter Scence Issues, Vo. 8, Issue 5, o, Setember 0 ISS (Onne: 694-084 www.ijcsi.org 75 onnear Robust Regresson Usng Kerne rnca Comonent Anayss and R-Estmators Anton Wbowo
More informationLinear Approximation with Regularization and Moving Least Squares
Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...
More informationPattern Recognition. Approximating class densities, Bayesian classifier, Errors in Biometric Systems
htt://.cubs.buffalo.edu attern Recognton Aromatng class denstes, Bayesan classfer, Errors n Bometrc Systems B. W. Slverman, Densty estmaton for statstcs and data analyss. London: Chaman and Hall, 986.
More informationA finite difference method for heat equation in the unbounded domain
Internatona Conerence on Advanced ectronc Scence and Technoogy (AST 6) A nte derence method or heat equaton n the unbounded doman a Quan Zheng and Xn Zhao Coege o Scence North Chna nversty o Technoogy
More informationThe Robustness of a Nash Equilibrium Simulation Model
8th World IMACS / MODSIM Congress, Carns, Australa 3-7 July 2009 htt://mssanz.org.au/modsm09 The Robustness of a Nash Equlbrum Smulaton Model Etaro Ayosh, Atsush Mak 2 and Takash Okamoto 3 Faculty of Scence
More informationUncertainty Specification and Propagation for Loss Estimation Using FOSM Methods
Uncertanty Specfcaton and Propagaton for Loss Estmaton Usng FOSM Methods J.W. Baer and C.A. Corne Dept. of Cv and Envronmenta Engneerng, Stanford Unversty, Stanford, CA 94305-400 Keywords: Sesmc, oss estmaton,
More informationSome Notes on Consumer Theory
Some Notes on Consumer Theory. Introducton In ths lecture we eamne the theory of dualty n the contet of consumer theory and ts use n the measurement of the benefts of rce and other changes. Dualty s not
More informationwe have E Y x t ( ( xl)) 1 ( xl), e a in I( Λ ) are as follows:
APPENDICES Aendx : the roof of Equaton (6 For j m n we have Smary from Equaton ( note that j '( ( ( j E Y x t ( ( x ( x a V ( ( x a ( ( x ( x b V ( ( x b V x e d ( abx ( ( x e a a bx ( x xe b a bx By usng
More informationBLIND SOURCE SEPARATION BASED ON THE FRACTIONAL FOURIER TRANSFORM
BLIND OURCE EARAION BAED ON HE FRACIONAL FOURIER RANFORM haron Karako-Eon (), Are Yeredor (),DavdMendovc () () Deartment o Eectrca Engneerng - ystems () Deartment o Eectrca Engneerng - hysca Eectroncs
More informationA DIMENSION-REDUCTION METHOD FOR STOCHASTIC ANALYSIS SECOND-MOMENT ANALYSIS
A DIMESIO-REDUCTIO METHOD FOR STOCHASTIC AALYSIS SECOD-MOMET AALYSIS S. Rahman Department of Mechanca Engneerng and Center for Computer-Aded Desgn The Unversty of Iowa Iowa Cty, IA 52245 June 2003 OUTLIE
More informationMulti-objective Optimal Block Transaction model based Transient Stability Evaluation
Proceedngs of the 5th WSEAS Internatona Conference on Aed Comuter Scence, Hangzhou, Chna, Ar 6-8, 006 (907-9) Mut-objectve Otma Boc Transacton mode based Transent Stabty Evauaton Ru Ma, Hongwen Yan Changsha
More informationCIS526: Machine Learning Lecture 3 (Sept 16, 2003) Linear Regression. Preparation help: Xiaoying Huang. x 1 θ 1 output... θ M x M
CIS56: achne Learnng Lecture 3 (Sept 6, 003) Preparaton help: Xaoyng Huang Lnear Regresson Lnear regresson can be represented by a functonal form: f(; θ) = θ 0 0 +θ + + θ = θ = 0 ote: 0 s a dummy attrbute
More informationNetworked Cooperative Distributed Model Predictive Control Based on State Observer
Apped Mathematcs, 6, 7, 48-64 ubshed Onne June 6 n ScRes. http://www.scrp.org/journa/am http://dx.do.org/.436/am.6.73 Networed Cooperatve Dstrbuted Mode redctve Contro Based on State Observer Ba Su, Yanan
More informationChapter 13: Multiple Regression
Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to
More informationSupervised Learning. Neural Networks and Back-Propagation Learning. Credit Assignment Problem. Feedforward Network. Adaptive System.
Part 7: Neura Networ & earnng /2/05 Superved earnng Neura Networ and Bac-Propagaton earnng Produce dered output for tranng nput Generaze reaonaby & appropratey to other nput Good exampe: pattern recognton
More informationA total variation approach
Denosng n dgtal radograhy: A total varaton aroach I. Froso M. Lucchese. A. Borghese htt://as-lab.ds.unm.t / 46 I. Froso, M. Lucchese,. A. Borghese Images are corruted by nose ) When measurement of some
More informationDiscriminating Fuzzy Preference Relations Based on Heuristic Possibilistic Clustering
Mutcrtera Orderng and ankng: Parta Orders, Ambgutes and Apped Issues Jan W. Owsńsk and aner Brüggemann, Edtors Dscrmnatng Fuzzy Preerence eatons Based on Heurstc Possbstc Custerng Dmtr A. Vattchenn Unted
More informationREDUCTION OF CORRELATION COMPUTATION IN THE PERMUTATION OF THE FREQUENCY DOMAIN ICA BY SELECTING DOAS ESTIMATED IN SUBARRAYS
15th European Sgna Processng Conference (EUSIPCO 27), Poznan, Poand, September 3-7, 27, copyrght by EURASIP REDUCTION OF CORRELATION COMPUTATION IN THE PERMUTATION OF THE FREQUENCY DOMAIN ICA BY SELECTING
More informationA Fluid-Based Model of Time-Limited TCP Flows 1
A Fud-Based Mode of Tme-Lmted TCP Fows Maro Barbera DIIT - Unversty of Catana V.e A. Dora 6 9525 Catana -Itay hone: +39 95 7382375 fax: +39 95 33828 mbarbera@.unct.t Afo Lombardo DIIT - Unversty of Catana
More informationMultiple Regression Analysis
Multle Regresson Analss Roland Szlág Ph.D. Assocate rofessor Correlaton descres the strength of a relatonsh, the degree to whch one varale s lnearl related to another Regresson shows us how to determne
More informationChapter 5: Repeated gravity measurements at Merapi volcano 45
Chater 5: Reeated gravty measurements at Mera vocano 45 5. Reeated Gravty Measurements at Mera Vocano As shown n 2.2 gravty changes n tme gve nformaton about subsurface mass mgraton. Such observatons try
More informationBias Term b in SVMs Again
Proceedngs of 2 th Euroean Symosum on Artfca Neura Networks,. 44-448, ESANN 2004, Bruges, Begum, 2004 Bas Term b n SVMs Agan Te Mng Huang, Vosav Kecman Schoo of Engneerng, The Unversty of Auckand, Auckand,
More informationDISTRIBUTED PROCESSING OVER ADAPTIVE NETWORKS. Cassio G. Lopes and Ali H. Sayed
DISTRIBUTED PROCESSIG OVER ADAPTIVE ETWORKS Casso G Lopes and A H Sayed Department of Eectrca Engneerng Unversty of Caforna Los Angees, CA, 995 Ema: {casso, sayed@eeucaedu ABSTRACT Dstrbuted adaptve agorthms
More informationComplete Variance Decomposition Methods. Cédric J. Sallaberry
Comlete Varance Decomoston Methods Cédrc J. allaberry enstvty Analyss y y [,,, ] [ y, y,, ] y ny s a vector o uncertan nuts s a vector o results s a comle uncton successon o derent codes, systems o de,
More informationPattern Classification (II) 杜俊
attern lassfcaton II 杜俊 junu@ustc.eu.cn Revew roalty & Statstcs Bayes theorem Ranom varales: screte vs. contnuous roalty struton: DF an DF Statstcs: mean, varance, moment arameter estmaton: MLE Informaton
More informationThe Leak Detection of Heating Pipe Based on Multi-Scale Correlation Algorithm of Wavelet
Sensors & Transducers Vo. 5 Speca Issue December 03 pp. 80-88 Sensors & Transducers 03 by IFSA http://www.sensorsporta.com The Lea Detecton of Heatng Ppe Based on ut-scae Correaton Agorthm of Waeet Xufang
More informationSupplementary Material for Spectral Clustering based on the graph p-laplacian
Sulementary Materal for Sectral Clusterng based on the grah -Lalacan Thomas Bühler and Matthas Hen Saarland Unversty, Saarbrücken, Germany {tb,hen}@csun-sbde May 009 Corrected verson, June 00 Abstract
More informationRESEARCH ARTICLE. Solving Polynomial Systems Using a Fast Adaptive Back Propagation-type Neural Network Algorithm
Juy 8, 6 8:57 Internatona Journa of Computer Mathematcs poynomas Internatona Journa of Computer Mathematcs Vo., No., Month, 9 RESEARCH ARTICLE Sovng Poynoma Systems Usng a Fast Adaptve Back Propagaton-type
More information+, where 0 x N - n. k k
CO 745, Mdterm Len Cabrera. A multle choce eam has questons, each of whch has ossble answers. A student nows the correct answer to n of these questons. For the remanng - n questons, he checs the answers
More informationTHERMODYNAMICS. Temperature
HERMODYNMICS hermodynamcs s the henomenologcal scence whch descrbes the behavor of macroscoc objects n terms of a small number of macroscoc arameters. s an examle, to descrbe a gas n terms of volume ressure
More informationHidden Markov Model Cheat Sheet
Hdden Markov Model Cheat Sheet (GIT ID: dc2f391536d67ed5847290d5250d4baae103487e) Ths document s a cheat sheet on Hdden Markov Models (HMMs). It resembles lecture notes, excet that t cuts to the chase
More informationThe line method combined with spectral chebyshev for space-time fractional diffusion equation
Apped and Computatona Mathematcs 014; 3(6): 330-336 Pubshed onne December 31, 014 (http://www.scencepubshnggroup.com/j/acm) do: 10.1164/j.acm.0140306.17 ISS: 3-5605 (Prnt); ISS: 3-5613 (Onne) The ne method
More informationCS 3710: Visual Recognition Classification and Detection. Adriana Kovashka Department of Computer Science January 13, 2015
CS 3710: Vsual Recognton Classfcaton and Detecton Adrana Kovashka Department of Computer Scence January 13, 2015 Plan for Today Vsual recognton bascs part 2: Classfcaton and detecton Adrana s research
More informationMixture of Gaussians Expectation Maximization (EM) Part 2
Mture of Gaussans Eectaton Mamaton EM Part 2 Most of the sldes are due to Chrstoher Bsho BCS Summer School Eeter 2003. The rest of the sldes are based on lecture notes by A. Ng Lmtatons of K-means Hard
More informationApplication of support vector machine in health monitoring of plate structures
Appcaton of support vector machne n heath montorng of pate structures *Satsh Satpa 1), Yogesh Khandare ), Sauvk Banerjee 3) and Anrban Guha 4) 1), ), 4) Department of Mechanca Engneerng, Indan Insttute
More informationWAVELET-BASED IMAGE COMPRESSION USING SUPPORT VECTOR MACHINE LEARNING AND ENCODING TECHNIQUES
WAVELE-BASED IMAGE COMPRESSION USING SUPPOR VECOR MACHINE LEARNING AND ENCODING ECHNIQUES Rakb Ahmed Gppsand Schoo of Computng and Informaton echnoogy Monash Unversty, Gppsand Campus Austraa. Rakb.Ahmed@nfotech.monash.edu.au
More informationIDENTIFICATION OF NONLINEAR SYSTEM VIA SVR OPTIMIZED BY PARTICLE SWARM ALGORITHM
Journa of Theoretca and Apped Informaton Technoogy th February 3. Vo. 48 No. 5-3 JATIT & LLS. A rghts reserved. ISSN: 99-8645 www.att.org E-ISSN: 87-395 IDENTIFICATION OF NONLINEAR SYSTEM VIA SVR OPTIMIZED
More informationHomework 10 Stat 547. Problem ) Z D!
Homework 0 Stat 547 Problem 74 Notaton: h s the hazard rate for the aneulod grou, h s the hazard rate for the dlod grou (a Log-rank test s erformed: H 0 : h (t = h (t Sgnfcance level α = 005 Test statstc
More informationA Three-Phase State Estimation in Unbalanced Distribution Networks with Switch Modelling
A Three-Phase State Estmaton n Unbaanced Dstrbuton Networks wth Swtch Modeng Ankur Majumdar Student Member, IEEE Dept of Eectrca and Eectronc Engneerng Impera Coege London London, UK ankurmajumdar@mperaacuk
More informationk p theory for bulk semiconductors
p theory for bu seconductors The attce perodc ndependent partce wave equaton s gven by p + V r + V p + δ H rψ ( r ) = εψ ( r ) (A) 4c In Eq. (A) V ( r ) s the effectve attce perodc potenta caused by the
More informationUsing T.O.M to Estimate Parameter of distributions that have not Single Exponential Family
IOSR Journal of Mathematcs IOSR-JM) ISSN: 2278-5728. Volume 3, Issue 3 Sep-Oct. 202), PP 44-48 www.osrjournals.org Usng T.O.M to Estmate Parameter of dstrbutons that have not Sngle Exponental Famly Jubran
More informationComparison of Outlier Detection Methods in Crossover Design Bioequivalence Studies
Journal of Pharmacy and Nutrton Scences, 01,, 16-170 16 Comarson of Outler Detecton Methods n Crossover Desgn Boequvalence Studes A. Rasheed 1,*, T. Ahmad,# and J.S. Sddq,# 1 Deartment of Research, Dow
More informationStatistical analysis using matlab. HY 439 Presented by: George Fortetsanakis
Statstcal analyss usng matlab HY 439 Presented by: George Fortetsanaks Roadmap Probablty dstrbutons Statstcal estmaton Fttng data to probablty dstrbutons Contnuous dstrbutons Contnuous random varable X
More informationResearch Article H Estimates for Discrete-Time Markovian Jump Linear Systems
Mathematca Probems n Engneerng Voume 213 Artce ID 945342 7 pages http://dxdoorg/11155/213/945342 Research Artce H Estmates for Dscrete-Tme Markovan Jump Lnear Systems Marco H Terra 1 Gdson Jesus 2 and
More informationJAB Chain. Long-tail claims development. ASTIN - September 2005 B.Verdier A. Klinger
JAB Chan Long-tal clams development ASTIN - September 2005 B.Verder A. Klnger Outlne Chan Ladder : comments A frst soluton: Munch Chan Ladder JAB Chan Chan Ladder: Comments Black lne: average pad to ncurred
More informationOn New Selection Procedures for Unequal Probability Sampling
Int. J. Oen Problems Comt. Math., Vol. 4, o. 1, March 011 ISS 1998-66; Coyrght ICSRS Publcaton, 011 www.-csrs.org On ew Selecton Procedures for Unequal Probablty Samlng Muhammad Qaser Shahbaz, Saman Shahbaz
More informationA Robust Method for Calculating the Correlation Coefficient
A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal
More informationLecture 10 Support Vector Machines II
Lecture 10 Support Vector Machnes II 22 February 2016 Taylor B. Arnold Yale Statstcs STAT 365/665 1/28 Notes: Problem 3 s posted and due ths upcomng Frday There was an early bug n the fake-test data; fxed
More informationResource Allocation and Decision Analysis (ECON 8010) Spring 2014 Foundations of Regression Analysis
Resource Allocaton and Decson Analss (ECON 800) Sprng 04 Foundatons of Regresson Analss Readng: Regresson Analss (ECON 800 Coursepak, Page 3) Defntons and Concepts: Regresson Analss statstcal technques
More informationLower bounds for the Crossing Number of the Cartesian Product of a Vertex-transitive Graph with a Cycle
Lower bounds for the Crossng Number of the Cartesan Product of a Vertex-transtve Graph wth a Cyce Junho Won MIT-PRIMES December 4, 013 Abstract. The mnmum number of crossngs for a drawngs of a gven graph
More informationA principal component analysis using SPSS for Multi-objective Decision Location Allocation Problem
Zpeng Zhang A prncpa component anayss usng SPSS for Mut-objectve Decson Locaton Aocaton Probem ZIPENG ZHANG Schoo of Management Scence and Engneerng Shandong Norma Unversty No.88 Cuture Rode, Jnan Cty,
More informationOptimization of JK Flip Flop Layout with Minimal Average Power of Consumption based on ACOR, Fuzzy-ACOR, GA, and Fuzzy-GA
Journa of mathematcs and computer Scence 4 (05) - 5 Optmzaton of JK Fp Fop Layout wth Mnma Average Power of Consumpton based on ACOR, Fuzzy-ACOR, GA, and Fuzzy-GA Farshd Kevanan *,, A Yekta *,, Nasser
More informationREAL-TIME IMPACT FORCE IDENTIFICATION OF CFRP LAMINATED PLATES USING SOUND WAVES
8 TH INTERNATIONAL CONERENCE ON COMPOSITE MATERIALS REAL-TIME IMPACT ORCE IDENTIICATION O CRP LAMINATED PLATES USING SOUND WAVES S. Atobe *, H. Kobayash, N. Hu 3 and H. ukunaga Department of Aerospace
More information