Spam Filtering Based on Support Vector Machines with Taguchi Method for Parameter Selection
|
|
- Conrad Fowler
- 5 years ago
- Views:
Transcription
1 E-mal Spam Flterng Based on Support Vector Machnes wth Taguch Method for Parameter Selecton We-Chh Hsu, Tsan-Yng Yu E-mal Spam Flterng Based on Support Vector Machnes wth Taguch Method for Parameter Selecton We-Chh Hsu 1, Tsan-Yng Yu 2,3 1 Department of Computer and Communcaton, Natonal Kaohsung Frst Unversty of Scence and Technology, Kaohsung, Tawan 2 Insttute of Engneerng Scence and Technology, Natonal Kaohsung Frst Unversty of Scence and Technology, Kaohsung, Tawan 3 Department of Electrcal Engneerng, Kao Yuan Unversty, Lu Chu, Tawan wechh@ccms.nkfust.edu.tw, yotnyg@gmal.com do: /jct.vol5.ssue8.9 Abstract Support Vector Machnes (SVM) s a powerful classfcaton technque n data mnng and has been successfully appled to many real-world applcatons. Parameter selecton of SVM wll affect classfcaton performance much durng tranng process. However, parameter selecton of SVM s usually dentfed by experence or grd search (GS). In ths study, we use Taguch method to make optmal approxmaton for the SVM-based E-mal Spam Flterng model. Sx real-world mal data sets are selected to demonstrate the effectveness and feasblty of the method. The results show that the Taguch method can fnd the effectve model wth hgh classfcaton accuracy. 1. Introducton Keywords: Support Vector Machnes (SVM), Taguch method, Grd search (GS) Spammng s the abuse of electronc messagng systems to send unsolcted bulk e-mals or to promote servces or products, whch are usually undesred. Spammng s economcally vable because advertsers have no operatng costs beyond the management of ther malng lsts. The sender cannot be specfed, because the sender of spammng has only temporary e-mal address and the reply of them s not reached to the orgnal sender. Therefore, undesred Emals to us have been ncreased everyday, so that, t s not easy to read an mportant e-mal. Early on the spam flterng black and whte lst was appled usually. Although fast and smple wth the characterstcs, but the drawback s that users have to update the spam mal flterng rules and mantan a black lst. Spam flterng based on the textual content of e-mal messages can be regarded as a specal case of text categorzaton, wth the categores beng spam and normal (non-spam). Contentbased flters can be dvded nto rule-based methods and probablstc methods. Rule-based methods such as Rpper [1-2], Boostng [3], Decson Tree [4], Rough Sets [5] and so on strongly dependent on the exstence of key terms, therefore, specfc terms can cause the falure of flterng. Methods based on probablty and statstcs such as K-Nearest Neghbor [5] and Support Vector Machne (SVM) [6] and so on. Besdes, the prevalng machne learnng method for spam message flterng s the Bayesan approach [7] used wth good results. SVM proposed by Vapnk [6] n 1995, has been wdely appled n many applcatons such as functon approxmaton, modelng, forecastng, optmzaton control...etc, and has yelded excellent performance. It s a statstcal theory to deal wth the dual categores of classfcaton and can fnd the best hyperplane to partton a sample space. Huang [8] demonstrate that the SVM-based model s very compettve to back-propagaton neural network (BPN), genetc programmng (GP) and decson tree n terms of classfcaton accuracy. Selecton of kernel functon s a pvotal factor whch decdes performance of SVM. RBF kernel functon penalty parameter C s most wdely used n SVM and few control parameters are requred. There are two parameters n and the kernel parameterγ. However, for the SVM-based model, ts classfcaton performance s senstve to the parameters of the model, thus, parameters selecton s very mportant. The optmzaton parameters ths functon (C, γ) wll make the SVM have the best performance. In spam flterng, the Bayesan algorthm n the mal system s very extensve. Compared wth Bayesan algorthm, f SVM s used wth lnear kernel functon or default 78
2 Journal of Convergence Informaton Technology Volume 5, Number 8, October 2010 parameters, the Bayesan algorthm wll be better than the accuracy of SVM. In order to enhance the accuracy of SVM, t s necessary to develop a search mechansm to tune the hyperparameters. Most of the prevous researches focus on the grd search (GS), pattern search based on prncples from desgn of experments (DOE) such as Staeln [9] and genetc algorthm (GA) [8, 10] to choose the parameters. GS s smple and easly mplemented, but t s very tme-consumng. DOE s lke GS but t reduces the searchng grd densty and can reduce the computatonal tme greatly. Although GA does not requre settng an ntal search range, t ntroduces some new parameters to control the GA searchng process, such as the populaton sze, generatons, and mutaton rate. The Taguch method [11], a robust desgn approach, uses many deas from statstcal expermental desgn for evaluatng and mplementng mprovements n products, processes, and equpment. The fundamental prncple s to mprove the qualty of a product by mnmzng the effect of the causes of varaton wthout elmnatng the causes. One of the major tools used n the Taguch method s orthogonal array (OA) to reduce the number of experments and obtan good expermental results. The parameters (C, γ) of SVM are regarded as control factors n OA. Experment s conducted through Multlevel-column OA after selectng the parameters of SVM. We verfy the classfcaton results and compared wth GS. As far as we know, ths maybe the frst attempt to ntroduce Taguch method to optmze the SVM for spam flterng models. The remander of ths paper s organzed as follows. In Secton 2, the SVM and Taguch method are descrbed brefly. Secton 3 presents mplementaton for our approach to classfy the spam e-mals. Secton 4 gves expermental results and dscusson. Fnally, the research results are summarzed and present future work. 2. The ntroducton of SVM and Taguch method The proposed approach s based on SVM and Taguch method. In ths secton, SVM and Taguch method are ntroduced brefly The bref descrpton of SVM The textual and non-textual features representng an emal, obtaned through the method mentoned prevously, are as the nput to the spam emal flterng algorthm. In the approach, the flterng algorthm s represented by SVM. SVM s a powerful supervsed learnng paradgm based on the structured rsk mnmzaton prncple from statstcal learnng theory, whch s currently placed among of the bestperformng classfers and have a unque ablty to handle extremely large feature spaces (such as text), precsely the area where most of the tradtonal technques fal due to the curse of the dmensonalty. SVM has been reported remarkable performance on text categorzaton task. In our evaluaton, we used the Lbrary for SVM [12] to buld SVM models. In the followng, we gve a bref ntroducton to the theory and mplementaton of SVM classfcaton algorthm. Consder the problem of separatng the set of tranng set vectors belongng to two separate classes n some feature space. Gven one set of tranng example vectors: ( x, y ),...( x, y ), x R, y { 1, 1} (1) 1 1 l l n + we try to separate the vectors wth a hyperplane so that ( w x) + b = 1 (2) y [( w x) + b] 1,( = 1, 2,..., l) (3) The hyperplane wth the largest margn s known as the optmal separatng hyperplane. It separates all vectors wthout error and the dstance between the closest vectors to the 79
3 E-mal Spam Flterng Based on Support Vector Machnes wth Taguch Method for Parameter Selecton We-Chh Hsu, Tsan-Yng Yu hyperplane s maxmal. The dstance s gven by 2 d ( w, b) = (4) w Hence the hyperplane that separates the data optmally s the one that mnmzes the followng equaton: 1 2 Mnmze w (5) 2 subject to the constrants of (4). To solve above problem, Lagrange multplers α are ntroduced. Let = 1,2,,l and defne w( ) = α y x l α (6) Wth Wolfe theory the problem can be transformed to ts dual problem: = 1 1 max W ( α ) = α w( α) w( α), s. t. α 0 2 (7) α = y 0 (8) Wth the optmal separatng hyperplane found, the decson functon can be wrtten as: f ( x) = ( w x) + (9) 0 b 0 Then the test data can be labeled wth label x) = sgn( f ( x)) = sgn(( w x) + ) (10) ( 0 b0 Tranng vectors that satsfy y [( w0 x) + b0 ] = 1 are termed support vectors, whch are always correspondng to nonzero α. The regon between the hyperplane through the support vectors on each sde s called the margn band. In the case of lnearly non-separable tranng data, by ntroducng slack varables the prmal problem can be rewrtten as: 1 2 Mn w + C ξ (11) 2 subject to y [( w x) + b] 1 ζ, ζ 0. Smlarly, we can get the correspondng dual problem maxw ( α) = s. t. C α 0, α 1 2 w( α) w( α), α y = 0 (12) 80
4 Journal of Convergence Informaton Technology Volume 5, Number 8, October 2010 Problems descrbed as n Equaton(11) and Equaton(12) are typcal quadratc optmzaton questons, and have been approached usng a varety of computatonal technques. Recent advances n optmzaton methods have made support vector learnng n large-scale tranng data possble. All the tranng vectors correspondng to nonzero α are called support vectors, whch form the boundares of the classes. The maxmal margn classfer can be generalzed to nonlnearly separable data va transformng nput vectors nto a hgher dmensonal feature space by a map functon ϕ, followed by a lnear separaton there. The expensve computaton of nner products can be reduced sgnfcantly by usng a sutable kernel functon K( x, x j ) = ( ϕ( x ), ϕ( x j )). We mplemented the SVM classfer usng the LIBSVM lbrary [12] and adopted radal bass 2 functon (RBF) defned as the kernel K( x =, x j ) exp γ x x j. In ths study, the RBF s used as the basc kernel functon of SVM. There are two parameters assocated wth the RBF kernels: C and γ. Vapnk found that a dfferent kernel functon of SVM has lttle effect on the performance but parameters of kernel functon are key factor The bref descrpton of Taguch Method In ths secton, we brefly ntroduce the basc concept of the structure and Taguch method. Taguch method s qute common n the desgn of ndustral experments [13-14]. Taguch method requres a sgnfcantly small number of experments compared wth other statstcal technques[15]. Although some nformaton s lost due to these two approxmatons, t s stll worth choosng ths approach, consderng the tme consumng nature. OA s a very mportant tool for Taguch method. Many desgned experments use matrces called OA for determnng whch combnatons of factor levels to use for each expermental run and for analyzng the data. An OA s a fractonal factoral matrx, whch assures a balanced comparson of levels of any factor or nteracton of factors. It s a matrx of numbers arranged n rows and columns where each row represents the level of the factors n each run, and each column represents a specfc factor that can be changed from each run. The array s called orthogonal because all columns can be evaluated ndependently of one another. The general symbol for m-level standard OA s ( m where n=m k number of expermental runs; k a postve nteger whch s greater than 1; m number of levels for each factor; n-1 number of columns n the OA. ) n 1 L n (13) The letter L comes from Latn, the dea of usng OA for expermental desgn havng been assocated wth Latn square desgns from the outset. The two-level standard OA whch are most often used n practce are L 4 (2 3 ), L 8 (2 7 ), L 16 (2 15 ), and L 32 (2 31 ). Table 1 shows an OA L 8 (2 7 ). 81
5 E-mal Spam Flterng Based on Support Vector Machnes wth Taguch Method for Parameter Selecton We-Chh Hsu, Tsan-Yng Yu Table 1. L 8 (2 7 ) Orthogonal array L 8 (2 7 ) Experment No cloumn The number to the left of each row s called the run number or experment number and runs from 1 to Implementaton In ths paper, the flow chart of e-mal spam flterng based on SVM wth Taguch method for parameter selecton s shown n Fgure 1. Frst stage s data pre-processng as depcted n Fgure 2. Vector space model s a text representng approach, whch s wdely used and has good performance n text categorzaton. In ts smple form, spam flterng can be recast as text Input messages Data Pre-processng Select Parameters Select OA Calculatng the expermental accuracy for each run Calculatng the effects of SVM parameters (C, γ) Optmal SVM parameters (C, γ) are obtaned based on prevous step Spam or Normal Fgure 1. The flow chart of e-mal spam flterng based on SVM wth Taguch method for parameter selecton Fgure 2. Data Pre-processng categorzaton task where the classes to be predcted are spam and normal. Therefore, Emal can be regarded as a vector space, whch s composed of a group of orthogonal key words. For each emal, ts textual porton was represented by a concatenaton of the subject lne and 82
6 Journal of Convergence Informaton Technology Volume 5, Number 8, October 2010 the body of the message. Due to the prevalence of html and bnary attachments n modern emal a degree of pre-processng s requred on messages to allow effectve feature extracton. Therefore, we adopt the followng data pre-processng steps: 1) If there exst HTML tags, then remove HTML tags. Then tokenzaton s the process of reducng a message to ts colloqual components. 2) To avod treatng forms of the same word as dfferent attrbutes, a lemmatzer was appled to the corpora to convert each word to ts base form (e.g., "got" becomes "get"). 3) The stoppng process s adopted to remove the hgh frequent words wth low content dscrmnatng power n an emal document such as "to", "a","and","t", etc. Removng these words wll save spaces for storng document contents and reduce tme taken durng the subsequent processes. We obtan word frequences and convert nto vectors. We ntroduce Taguch method to our approach. In content-based spam flterng performance analyss, a commonly used evaluaton crtera measurng the effcency of the classfcaton s accuracy (Acc). It s regarded as response varable, defned as: A + D Acc = (14) N where N s the number of all messages; A s as spam and the actual system to determne the number of spam; and D that the actual system for normal mal and e-mal to determne the number of normal. Table 2. Descrpton of data sets Orgnal Our method Data set Non-Spam Spam Non-Spam Spam enronspam 16,545 17, lngspam 2, , PU PU PU3 2,313 1, PUA Table 3. Experment set-up and data for L 16 (8 8 2) Orgnal Columns 1, 2, 4 13, 6, 1 Exp. Modfed columns No. 1 2 Acc Factor Log 2 (c) Log 2 (γ) PU1 PU2 PU3 PUA lngspam enronspam
7 E-mal Spam Flterng Based on Support Vector Machnes wth Taguch Method for Parameter Selecton We-Chh Hsu, Tsan-Yng Yu Table 4. L 16 (8 8 2) OA and experment data PU1 PU2 PU3 PUA lngspam enronspam Factor Log 2 (C) Log 2 (γ) Log 2 (C) Log 2 (γ) Log 2 (C) Log 2 (γ) Log 2 (C) Log 2 (γ) Log 2 (C) Log 2 (γ) Log 2 (C) Log 2 (γ) Max Mn Effect In order to reduce the number of experments and the cost of desgn, we have to choose approprate OA by numbers of control factors and levels. To explan how to employ OA to obtan the soluton, on the other hand as the search scope s suggested by Ln [12] and we expand to dfferent combnatons of parameters C and γ wth 8 levels: log 2 (C) = (-15, -11, -6, -2, 3, 7, 12, 16) and log 2 (γ) = (-15, -11, -6, -2, 3, 7, 12, 16) to fnd the best combnaton. In thswork, both of the factor log 2 (C), log 2 (γ) are set at eght levels. Seven degrees of freedom (d.f.) are requred for each factor. Consder that 14 d.f. are requred n total, an OA type L 16 (2 15 ) wth 16 trals and 15 d.f., as ndcated n left sde Table 3, s adopted. A converson of the L 16 array of two levels to one multlevel wth 8 levels had to be performed to accommodate two factors log 2 (C), log 2 (γ) wth 8 levels. Ths modfcaton of the OA should be planned n such a way that respects the d.f. of the L 16. In general, three man concepts were used n the orthogonal arrays theory [16]. 1. Balance, for each factor the levels occur equally often. 2. Estmablty, every parameter could be capable of beng estmated. 3. Orthogonalty, a term whch mples that t s easy to extract and separate out the effect of dfferent factors equally. Multlevel factors could be created by the approprate multlevel columns n two-level arrays. Ths s generally acheved at expense of 3 columns whch are replaced by a new column whose levels drectly correspond to every level-combnaton of the orgnal 2 columns. The only requrement for the creaton of multlevel columns n ths way s that four nteracton columns must exst for the 3 sacrfced columns; these are deleted. Consequently, only one two-level column s leaved to reman after converson and L 16 (8 8 2) are acheved. In order to verfy whether the arthmetc s vald or not, we empoly 5-fold cross valdaton for our experment. 5-fold cross valdaton s to separate e-mals nto 5 parts. We make use of the 4 parts for tranng, and the remanng for testng. The procedure loops 5 tmes, so every part has been tested. Fnally, the average of tests values s used as the result of test for evaluaton. Each run of L 16 (8 8 2) wll proceed 5-fold cross valdaton. The accuracy for each run and the average accuracy for each level and each factor need to be evaluated. We pck the level wth maxmum accuracy for each factor. Therefore, we can obtan approxmaton results. 4. Experment results and dscusson In our test, the program runs wth LIBSVM toolbox provde by Ln [12] on an IBM compatble PC wth AMD Athlon CPU runnng at 1.8 GHz wth 1GB RAM. Sx publc data sets have been used n ths study. The experments were conducted on the PU corpora, the lngspam corpus and enronspam corpora. The four PU corpora, PU1, PU2, PU3 and PUA, respectvely, have been made publcly avalable by Androutsopoulos et al. They are encrypted data sets n order to promote standard benchmarks. Lngspam s a mxture of
8 Journal of Convergence Informaton Technology Volume 5, Number 8, October 2010 spam messages and 2412 messages sent va the Lngust lst, a moderated (hence, spam-free) lst about the professon and scence of lngustcs. Attachments, HTML tags, and duplcate spam messages receved on the same day are not ncluded. The enronspam corpus 1, whch are sx nonencoded data sets, contans ham messages of partcular users and fresh spam messages and ncludes spam messages from varous sources. We mx ths enronspam and take 500 normal messages and 500 spam messages randomly. Table 2 shows the summary of the data sets. Regardng messages n PU1, PU2 and the PUA are not many, so all spam and non-spam messages are put n our test. Data set Table 5. Results for dfferent OAs GS(8 8) L16(8 8 2) Log(C) Log(γ) Acc(%) Log(C) Log(γ) Acc(%) Acc. Dff.(%) pu pu pu pua lngspam enronspam Avg (a) GS(16 16) L32( ) Data set Acc. Log(C) Log(γ) Acc(%) Log(C) Log(γ) Acc(%) Dff.(%) pu pu pu pua lngspam enronspam Avg (b) GS(32 32) L 64 ( ) Data set Acc. Log(C) Log(γ) Acc(%) Log(C) Log(γ) Acc(%) Dff.(%) pu pu pu pua lngspam enronspam Avg Experment set-up and data for L 16 (8 8 2) s shown n Table 3. In ths table, the converson of L 16 (8 8 2) from L 16 (2 15 ) stll keep orthogonal. It ndcates that the accuracy of SVM wll become worse wthout careful selecton for parameters C and γ. We lst accuracy averages of both parameters log 2 (C) and log 2 (γ) for every level n dfferent data sets and evaluate effectve of control factors for all levels as llustrated n Table 4. Here accuracy s desrable as larger as n possble. The maxmum of both parameters log 2 (C) and log 2 (γ) accuracy average for each level each data set are marked. The dfference between maxmum accuracy and mnmum 1 The Enron-Spam datasets are avalable from and n both raw and pre-processed form. Lng-Spam and the pu corpora are also avalable from the same addresses 85
9 E-mal Spam Flterng Based on Support Vector Machnes wth Taguch Method for Parameter Selecton We-Chh Hsu, Tsan-Yng Yu accuracy of man effect for parameters log 2 (C) and log 2 (γ) mples the mpact for accuracy. By observng the effectve and varance of control factor log 2 (γ) and log 2 (C) for all level, The dfference of parameters log 2 (γ) s lager than the one of parameters log 2 (γ). It means that parameter γ s more sgnfcant than parameter C for all data sets. The experment of both methods used (a) (b) (c) (d) (e) (f) Fgure 3. The contour plots of GS on C = (2-15, 2-14, 2-13, 2-12,..., 2 15, 2-16 ) and γ = (2-1 5, 2-14, 2-13, 2-12,..., 2 15, 2-16 ) for (a) PU1 (b) PU2 (c) PU3 (d) PUA (e) enronspam (f) lngspam dentcal tranng and testng sets wth 5-fold cross valdaton. The average classfcaton accuracy of 5-fold cross valdaton of both methods for In other data sets, the average accuracy for Taguch method s close to the results for GS but not good enough. Furthermore, we apply Taguch method wth more levels OAs as depcted n Table 5(b)(c). Ths mprovement s sgnfcant between Table 5(a) and (b). However, the mprovement s lttle between Table 5(b) 86
10 Journal of Convergence Informaton Technology Volume 5, Number 8, October 2010 and (c). Taguch approach wth more the number of levels has more effectve detectve ponts, so the accuracy wll get hgher. Meanwhle, the dfference n accuracy between GS and our proposed method wll decrease. The comparson of both methods s based on the same levels n Accuracy 100% 80% 60% 40% 20% 0% PU1 PU2 PU3 PUA lngspam eronspam Nave Bayesan 84.99% 90.93% 89.90% 90.19% 84.57% 78.27% SVM (Lnear) 78.78% 81.39% 83.18% 83.96% 89.70% 86.17% SVM (Taguch Method L32) 92.99% 92.56% 93.90% 91.30% 98.70% 94.97% SVM (GS 32 32) 94.27% 94.17% 95.60% 92.20% 98.89% 95.47% Fgure 4. Accuracy for dfferent methods ths experment. Fgure 3 s avalable by GS on C = (2-15, 2-14, 2-13, 2-12,..., 2 15, 2-16 ) and γ = (2-15, 2-14, 2-13, 2-12,..., 2 15, 2-16 ) for each data set. Hgher accuraces concentrated n the lower rght corner of the contour graph. These contrbutons are smlar for all data sets. Compared wth Taguch method, SVM wth lnear kernel, GS and Naїve Bayes algorthm for dfferent data sets, the results of our confrm test are shown n Fgure 4. The results ndcate to be set up for SVM wth lnear kernel but the accuracy wll lower than that of Naїve Bayes algorthms and our proposed method. As for tme complexty, GS requred searchng and computng = 1024 tmes but our proposed method need only 64 tmes. Our approach s 15 tmes faster and accuracy of our proposed method s very close to that of GS. The expermental results show that our proposed method can select good parameters for SVM wth kernel RBF and the accuracy s very close to that of GS. 5. Conclusons and future work Our proposed approach based on Taguch method does not lke other approxmaton methods or heurstcs may cause exhaustve parameter searches. On the other hand, our proposed approach sometmes may obtan approxmaton results but not optmal. However, compared wth much computatonal tme to fnd the optmal parameter values by the grd-search, t s worth for our methods to obtan approxmaton results at expense of lttle accuracy. From above experments, approprate OA could acheve hgh accuracy but hgh multlevel OA make lttle mprovement. In order to acheve approprate multlevel-column OA, we convert from 2-level OA and stll keep multlevel-column OA orthogonal. In our method the parameter selecton by orthogonal table wll obtan hgh accuracy. If we would lke to obtan hgher accuracy, we could extend OA L 64 to an OA such as L 128 to promote the accuracy. 6. References [1] W. W. Cohen, "Fast effectve rule nducton," n Proceedngs of the Twelfth Internatonal Conference on Machne Learnng, 1995, pp [2] W. C. Wllam, "Learnng rules that classfy e-mal," n Proceedngs of the 1996 AAAI Sprng Sympo-sum n Informaton Access, 1996, pp
11 E-mal Spam Flterng Based on Support Vector Machnes wth Taguch Method for Parameter Selecton We-Chh Hsu, Tsan-Yng Yu [3] I. Androutsopoulos, et al., Learnng to flter unsolcted commercal e-mal: "DEMOKRITOS", Natonal Center for Scentfc Research, [4] X. Carreras and L. Marquez, "Boostng trees for ant-spam emal flterng," n Proceedngs of RANLP-01,4th Internatonal Conference on Recent Advances n Natural Language Processng, Tzgov Chark, BG, [5] I. Androutsopoulos, et al., "Learnng to flter spam e-mal: A comparson of a nave bayesan and a memory-based approach," [6] V. N. Vapnk, The Nature of Statstcal Learnng Theory: New York: Sprnger-Verlag, [7] J. Provost, "Nave-bayes vs. rule-learnng n classfcaton of emal. The Unversty of Texas at Austn," Artfcal Intellgence Lab. Techncal Report AI-TR , [8] C. L. Huang and C. J. Wang, "A GA-based feature selecton and parameters optmzatonfor support vector machnes," Expert Systems Wth Applcatons, vol. 31, pp , [9] C. Staeln, "Parameter selecton for support vector machnes," Hewlett-Packard Company, Tech. Rep. HPL R1, [10] T. Howley and M. G. Madden, "The genetc kernel support vector machne: Descrpton and evaluaton," Artfcal Intellgence Revew, vol. 24, pp , [11] Taguch and S. Chowdhury, Robust engneerng: McGraw-Hll Professonal, [12] C. C. Chang and C. J. Ln, "LIBSVM -- A Lbrary for Support Vector Machnes," [13] G. Taguch, Introducton to qualty engneerng: Asan Productvty Organzaton Tokyo, [14] M. Phadke, Qualty engneerng usng robust desgn: Prentce Hall PTR Upper Saddle Rver, NJ, USA, [15] D. C. Montgomery, Desgn and analyss of experments. New York: Wley, [16] N. Logothets and H. P. Wynn, Qualty through desgn: expermental desgn, off-lne qualty control, and Taguch's contrbuton. Oxford: Clarendon Press,
Support Vector Machines
/9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.
More informationThe Research of Support Vector Machine in Agricultural Data Classification
The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou
More informationLearning the Kernel Parameters in Kernel Minimum Distance Classifier
Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department
More informationSupport Vector Machines
Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned
More informationClassification / Regression Support Vector Machines
Classfcaton / Regresson Support Vector Machnes Jeff Howbert Introducton to Machne Learnng Wnter 04 Topcs SVM classfers for lnearly separable classes SVM classfers for non-lnearly separable classes SVM
More informationEdge Detection in Noisy Images Using the Support Vector Machines
Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona
More informationCluster Analysis of Electrical Behavior
Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School
More informationClassifier Selection Based on Data Complexity Measures *
Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.
More informationA Binarization Algorithm specialized on Document Images and Photos
A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a
More informationSmoothing Spline ANOVA for variable screening
Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory
More informationParallelism for Nested Loops with Non-uniform and Flow Dependences
Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr
More informationTsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance
Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for
More informationCHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION
48 CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION 3.1 INTRODUCTION The raw mcroarray data s bascally an mage wth dfferent colors ndcatng hybrdzaton (Xue
More informationOutline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1
4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:
More informationDetermining the Optimal Bandwidth Based on Multi-criterion Fusion
Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn
More informationCS246: Mining Massive Datasets Jure Leskovec, Stanford University
CS46: Mnng Massve Datasets Jure Leskovec, Stanford Unversty http://cs46.stanford.edu /19/013 Jure Leskovec, Stanford CS46: Mnng Massve Datasets, http://cs46.stanford.edu Perceptron: y = sgn( x Ho to fnd
More informationTerm Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task
Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto
More informationAn Optimal Algorithm for Prufer Codes *
J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,
More informationProblem Definitions and Evaluation Criteria for Computational Expensive Optimization
Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty
More informationMachine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law)
Machne Learnng Support Vector Machnes (contans materal adapted from talks by Constantn F. Alfers & Ioanns Tsamardnos, and Martn Law) Bryan Pardo, Machne Learnng: EECS 349 Fall 2014 Support Vector Machnes
More informationA Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines
A Modfed Medan Flter for the Removal of Impulse Nose Based on the Support Vector Machnes H. GOMEZ-MORENO, S. MALDONADO-BASCON, F. LOPEZ-FERRERAS, M. UTRILLA- MANSO AND P. GIL-JIMENEZ Departamento de Teoría
More informationBOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET
1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School
More informationSupport Vector Machines. CS534 - Machine Learning
Support Vector Machnes CS534 - Machne Learnng Perceptron Revsted: Lnear Separators Bnar classfcaton can be veed as the task of separatng classes n feature space: b > 0 b 0 b < 0 f() sgn( b) Lnear Separators
More informationMeta-heuristics for Multidimensional Knapsack Problems
2012 4th Internatonal Conference on Computer Research and Development IPCSIT vol.39 (2012) (2012) IACSIT Press, Sngapore Meta-heurstcs for Multdmensonal Knapsack Problems Zhbao Man + Computer Scence Department,
More informationSolving two-person zero-sum game by Matlab
Appled Mechancs and Materals Onlne: 2011-02-02 ISSN: 1662-7482, Vols. 50-51, pp 262-265 do:10.4028/www.scentfc.net/amm.50-51.262 2011 Trans Tech Publcatons, Swtzerland Solvng two-person zero-sum game by
More informationAnnouncements. Supervised Learning
Announcements See Chapter 5 of Duda, Hart, and Stork. Tutoral by Burge lnked to on web page. Supervsed Learnng Classfcaton wth labeled eamples. Images vectors n hgh-d space. Supervsed Learnng Labeled eamples
More informationContent Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers
IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth
More information12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification
Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero
More informationFace Recognition Based on SVM and 2DPCA
Vol. 4, o. 3, September, 2011 Face Recognton Based on SVM and 2DPCA Tha Hoang Le, Len Bu Faculty of Informaton Technology, HCMC Unversty of Scence Faculty of Informaton Scences and Engneerng, Unversty
More informationNUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS
ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana
More informationIncremental Learning with Support Vector Machines and Fuzzy Set Theory
The 25th Workshop on Combnatoral Mathematcs and Computaton Theory Incremental Learnng wth Support Vector Machnes and Fuzzy Set Theory Yu-Mng Chuang 1 and Cha-Hwa Ln 2* 1 Department of Computer Scence and
More informationAn Anti-Noise Text Categorization Method based on Support Vector Machines *
An Ant-Nose Text ategorzaton Method based on Support Vector Machnes * hen Ln, Huang Je and Gong Zheng-Hu School of omputer Scence, Natonal Unversty of Defense Technology, hangsha, 410073, hna chenln@nudt.edu.cn,
More informationAPPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT
3. - 5. 5., Brno, Czech Republc, EU APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT Abstract Josef TOŠENOVSKÝ ) Lenka MONSPORTOVÁ ) Flp TOŠENOVSKÝ
More informationFeature Reduction and Selection
Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components
More informationAn Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices
Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal
More informationMachine Learning 9. week
Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below
More informationSubspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;
Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features
More informationX- Chart Using ANOM Approach
ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are
More informationAn Entropy-Based Approach to Integrated Information Needs Assessment
Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology
More informationUsing Neural Networks and Support Vector Machines in Data Mining
Usng eural etworks and Support Vector Machnes n Data Mnng RICHARD A. WASIOWSKI Computer Scence Department Calforna State Unversty Domnguez Hlls Carson, CA 90747 USA Abstract: - Multvarate data analyss
More informationSkew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach
Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research
More informationA Facet Generation Procedure. for solving 0/1 integer programs
A Facet Generaton Procedure for solvng 0/ nteger programs by Gyana R. Parja IBM Corporaton, Poughkeepse, NY 260 Radu Gaddov Emery Worldwde Arlnes, Vandala, Oho 45377 and Wlbert E. Wlhelm Teas A&M Unversty,
More informationA mathematical programming approach to the analysis, design and scheduling of offshore oilfields
17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 A mathematcal programmng approach to the analyss, desgn and
More informationLearning-Based Top-N Selection Query Evaluation over Relational Databases
Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **
More informationWishing you all a Total Quality New Year!
Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma
More informationCS 534: Computer Vision Model Fitting
CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust
More informationMULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION
MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and
More informationSum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints
Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan
More informationRelated-Mode Attacks on CTR Encryption Mode
Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory
More informationNetwork Intrusion Detection Based on PSO-SVM
TELKOMNIKA Indonesan Journal of Electrcal Engneerng Vol.1, No., February 014, pp. 150 ~ 1508 DOI: http://dx.do.org/10.11591/telkomnka.v1.386 150 Network Intruson Detecton Based on PSO-SVM Changsheng Xang*
More informationFEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur
FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents
More informationA New Approach For the Ranking of Fuzzy Sets With Different Heights
New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays
More informationCompiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz
Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster
More informationA Fast Content-Based Multimedia Retrieval Technique Using Compressed Data
A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,
More informationDiscriminative classifiers for object classification. Last time
Dscrmnatve classfers for object classfcaton Thursday, Nov 12 Krsten Grauman UT Austn Last tme Supervsed classfcaton Loss and rsk, kbayes rule Skn color detecton example Sldng ndo detecton Classfers, boostng
More informationSVM-based Learning for Multiple Model Estimation
SVM-based Learnng for Multple Model Estmaton Vladmr Cherkassky and Yunqan Ma Department of Electrcal and Computer Engneerng Unversty of Mnnesota Mnneapols, MN 55455 {cherkass,myq}@ece.umn.edu Abstract:
More informationA Fast Visual Tracking Algorithm Based on Circle Pixels Matching
A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng
More informationProper Choice of Data Used for the Estimation of Datum Transformation Parameters
Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and
More informationFace Recognition University at Buffalo CSE666 Lecture Slides Resources:
Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural
More informationImprovement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration
Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,
More informationAnalysis of Continuous Beams in General
Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,
More informationProgramming in Fortran 90 : 2017/2018
Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values
More informationComplex System Reliability Evaluation using Support Vector Machine for Incomplete Data-set
Internatonal Journal of Performablty Engneerng, Vol. 7, No. 1, January 2010, pp.32-42. RAMS Consultants Prnted n Inda Complex System Relablty Evaluaton usng Support Vector Machne for Incomplete Data-set
More informationConditional Speculative Decimal Addition*
Condtonal Speculatve Decmal Addton Alvaro Vazquez and Elsardo Antelo Dep. of Electronc and Computer Engneerng Unv. of Santago de Compostela, Span Ths work was supported n part by Xunta de Galca under grant
More informationJournal of Chemical and Pharmaceutical Research, 2014, 6(6): Research Article. A selective ensemble classification method on microarray data
Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(6):2860-2866 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A selectve ensemble classfcaton method on mcroarray
More informationHigh-Boost Mesh Filtering for 3-D Shape Enhancement
Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,
More informationFeature Selection as an Improving Step for Decision Tree Construction
2009 Internatonal Conference on Machne Learnng and Computng IPCSIT vol.3 (2011) (2011) IACSIT Press, Sngapore Feature Selecton as an Improvng Step for Decson Tree Constructon Mahd Esmael 1, Fazekas Gabor
More informationCLASSIFICATION OF ULTRASONIC SIGNALS
The 8 th Internatonal Conference of the Slovenan Socety for Non-Destructve Testng»Applcaton of Contemporary Non-Destructve Testng n Engneerng«September -3, 5, Portorož, Slovena, pp. 7-33 CLASSIFICATION
More informationInvestigating the Performance of Naïve- Bayes Classifiers and K- Nearest Neighbor Classifiers
Journal of Convergence Informaton Technology Volume 5, Number 2, Aprl 2010 Investgatng the Performance of Naïve- Bayes Classfers and K- Nearest Neghbor Classfers Mohammed J. Islam *, Q. M. Jonathan Wu,
More informationLecture 5: Multilayer Perceptrons
Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented
More informationLoad-Balanced Anycast Routing
Load-Balanced Anycast Routng Chng-Yu Ln, Jung-Hua Lo, and Sy-Yen Kuo Department of Electrcal Engneerng atonal Tawan Unversty, Tape, Tawan sykuo@cc.ee.ntu.edu.tw Abstract For fault-tolerance and load-balance
More informationEfficient Text Classification by Weighted Proximal SVM *
Effcent ext Classfcaton by Weghted Proxmal SVM * Dong Zhuang 1, Benyu Zhang, Qang Yang 3, Jun Yan 4, Zheng Chen, Yng Chen 1 1 Computer Scence and Engneerng, Bejng Insttute of echnology, Bejng 100081, Chna
More informationA Robust Method for Estimating the Fundamental Matrix
Proc. VIIth Dgtal Image Computng: Technques and Applcatons, Sun C., Talbot H., Ourseln S. and Adraansen T. (Eds.), 0- Dec. 003, Sydney A Robust Method for Estmatng the Fundamental Matrx C.L. Feng and Y.S.
More informationOutline. Type of Machine Learning. Examples of Application. Unsupervised Learning
Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton
More informationImplementation Naïve Bayes Algorithm for Student Classification Based on Graduation Status
Internatonal Journal of Appled Busness and Informaton Systems ISSN: 2597-8993 Vol 1, No 2, September 2017, pp. 6-12 6 Implementaton Naïve Bayes Algorthm for Student Classfcaton Based on Graduaton Status
More informationS1 Note. Basis functions.
S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type
More informationAn Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation
17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed
More informationUB at GeoCLEF Department of Geography Abstract
UB at GeoCLEF 2006 Mguel E. Ruz (1), Stuart Shapro (2), June Abbas (1), Slva B. Southwck (1) and Davd Mark (3) State Unversty of New York at Buffalo (1) Department of Lbrary and Informaton Studes (2) Department
More informationUser Authentication Based On Behavioral Mouse Dynamics Biometrics
User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA
More informationGA-Based Learning Algorithms to Identify Fuzzy Rules for Fuzzy Neural Networks
Seventh Internatonal Conference on Intellgent Systems Desgn and Applcatons GA-Based Learnng Algorthms to Identfy Fuzzy Rules for Fuzzy Neural Networks K Almejall, K Dahal, Member IEEE, and A Hossan, Member
More informationHuman Face Recognition Using Generalized. Kernel Fisher Discriminant
Human Face Recognton Usng Generalzed Kernel Fsher Dscrmnant ng-yu Sun,2 De-Shuang Huang Ln Guo. Insttute of Intellgent Machnes, Chnese Academy of Scences, P.O.ox 30, Hefe, Anhu, Chna. 2. Department of
More informationClassification of Face Images Based on Gender using Dimensionality Reduction Techniques and SVM
Classfcaton of Face Images Based on Gender usng Dmensonalty Reducton Technques and SVM Fahm Mannan 260 266 294 School of Computer Scence McGll Unversty Abstract Ths report presents gender classfcaton based
More informationSupport Vector classifiers for Land Cover Classification
Map Inda 2003 Image Processng & Interpretaton Support Vector classfers for Land Cover Classfcaton Mahesh Pal Paul M. Mather Lecturer, department of Cvl engneerng Prof., School of geography Natonal Insttute
More informationy and the total sum of
Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton
More informationDetection of an Object by using Principal Component Analysis
Detecton of an Object by usng Prncpal Component Analyss 1. G. Nagaven, 2. Dr. T. Sreenvasulu Reddy 1. M.Tech, Department of EEE, SVUCE, Trupath, Inda. 2. Assoc. Professor, Department of ECE, SVUCE, Trupath,
More informationVirtual Machine Migration based on Trust Measurement of Computer Node
Appled Mechancs and Materals Onlne: 2014-04-04 ISSN: 1662-7482, Vols. 536-537, pp 678-682 do:10.4028/www.scentfc.net/amm.536-537.678 2014 Trans Tech Publcatons, Swtzerland Vrtual Machne Mgraton based on
More informationMachine Learning. Topic 6: Clustering
Machne Learnng Topc 6: lusterng lusterng Groupng data nto (hopefully useful) sets. Thngs on the left Thngs on the rght Applcatons of lusterng Hypothess Generaton lusters mght suggest natural groups. Hypothess
More informationThe Study of Remote Sensing Image Classification Based on Support Vector Machine
Sensors & Transducers 03 by IFSA http://www.sensorsportal.com The Study of Remote Sensng Image Classfcaton Based on Support Vector Machne, ZHANG Jan-Hua Key Research Insttute of Yellow Rver Cvlzaton and
More informationAn Image Fusion Approach Based on Segmentation Region
Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua
More informationMachine Learning: Algorithms and Applications
14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of
More informationData Representation in Digital Design, a Single Conversion Equation and a Formal Languages Approach
Data Representaton n Dgtal Desgn, a Sngle Converson Equaton and a Formal Languages Approach Hassan Farhat Unversty of Nebraska at Omaha Abstract- In the study of data representaton n dgtal desgn and computer
More informationR s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes
SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges
More informationTN348: Openlab Module - Colocalization
TN348: Openlab Module - Colocalzaton Topc The Colocalzaton module provdes the faclty to vsualze and quantfy colocalzaton between pars of mages. The Colocalzaton wndow contans a prevew of the two mages
More informationConcurrent Apriori Data Mining Algorithms
Concurrent Apror Data Mnng Algorthms Vassl Halatchev Department of Electrcal Engneerng and Computer Scence York Unversty, Toronto October 8, 2015 Outlne Why t s mportant Introducton to Assocaton Rule Mnng
More informationParallel matrix-vector multiplication
Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more
More informationMathematics 256 a course in differential equations for engineering students
Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the
More informationA MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS
Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung
More informationSimulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010
Smulaton: Solvng Dynamc Models ABE 5646 Week Chapter 2, Sprng 200 Week Descrpton Readng Materal Mar 5- Mar 9 Evaluatng [Crop] Models Comparng a model wth data - Graphcal, errors - Measures of agreement
More informationOn Some Entertaining Applications of the Concept of Set in Computer Science Course
On Some Entertanng Applcatons of the Concept of Set n Computer Scence Course Krasmr Yordzhev *, Hrstna Kostadnova ** * Assocate Professor Krasmr Yordzhev, Ph.D., Faculty of Mathematcs and Natural Scences,
More informationPerformance Evaluation of Information Retrieval Systems
Why System Evaluaton? Performance Evaluaton of Informaton Retreval Systems Many sldes n ths secton are adapted from Prof. Joydeep Ghosh (UT ECE) who n turn adapted them from Prof. Dk Lee (Unv. of Scence
More information